Interactive Evolutionary Computation Hideyuki TAKAGI Kyushu University [email protected]...
-
Upload
clara-bourke -
Category
Documents
-
view
218 -
download
2
Transcript of Interactive Evolutionary Computation Hideyuki TAKAGI Kyushu University [email protected]...
Interactive Evolutionary Computation
Hideyuki TAKAGI
Kyushu Universitytakagidesignkyushu-uacjp
http wwwdesignkyushu-uacjp~takagi
Part 1 Humanized Computational Intelligence
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
neuron model (1943)
Hebbian law (1949)
Perceptron (1958) Adaline
(1960)
Fuzzy Logic (1965)
Fraser (1950s)
GAHolland (1960s)
EPFogel (1960s)
ESRechenberg (1960s-70s)
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
era of NN FS and EC
era of NN+FS+EC
Computational Intelligence in the 20th Century
Cooperation ofComputational Intelligence
EC
KE FS
NN
Powerful cooperative technologies were developed in the end of 20th Century
What Comes Next
System optimization based on human evaluation
Computer support system for creativity psychological and physical satisfaction
EC
KE FS
NN
+computer real human
Knowledge KANSEI
reasoning
expression
acquisition
associative memory
learning
intuition
preference
subjective evaluation
perception
cognition
etc etc
Two DifferentHuman Capabilities
explicitknowledge
experience
preference
computationalintelligence
optimization
auto-design
implicitknowledge
1st Generation
2nd Generation
3rd Generation
Importance of Human Factors withComputational Intelligence
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Part 1 Humanized Computational Intelligence
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
neuron model (1943)
Hebbian law (1949)
Perceptron (1958) Adaline
(1960)
Fuzzy Logic (1965)
Fraser (1950s)
GAHolland (1960s)
EPFogel (1960s)
ESRechenberg (1960s-70s)
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
era of NN FS and EC
era of NN+FS+EC
Computational Intelligence in the 20th Century
Cooperation ofComputational Intelligence
EC
KE FS
NN
Powerful cooperative technologies were developed in the end of 20th Century
What Comes Next
System optimization based on human evaluation
Computer support system for creativity psychological and physical satisfaction
EC
KE FS
NN
+computer real human
Knowledge KANSEI
reasoning
expression
acquisition
associative memory
learning
intuition
preference
subjective evaluation
perception
cognition
etc etc
Two DifferentHuman Capabilities
explicitknowledge
experience
preference
computationalintelligence
optimization
auto-design
implicitknowledge
1st Generation
2nd Generation
3rd Generation
Importance of Human Factors withComputational Intelligence
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
neuron model (1943)
Hebbian law (1949)
Perceptron (1958) Adaline
(1960)
Fuzzy Logic (1965)
Fraser (1950s)
GAHolland (1960s)
EPFogel (1960s)
ESRechenberg (1960s-70s)
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
era of NN FS and EC
era of NN+FS+EC
Computational Intelligence in the 20th Century
Cooperation ofComputational Intelligence
EC
KE FS
NN
Powerful cooperative technologies were developed in the end of 20th Century
What Comes Next
System optimization based on human evaluation
Computer support system for creativity psychological and physical satisfaction
EC
KE FS
NN
+computer real human
Knowledge KANSEI
reasoning
expression
acquisition
associative memory
learning
intuition
preference
subjective evaluation
perception
cognition
etc etc
Two DifferentHuman Capabilities
explicitknowledge
experience
preference
computationalintelligence
optimization
auto-design
implicitknowledge
1st Generation
2nd Generation
3rd Generation
Importance of Human Factors withComputational Intelligence
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Cooperation ofComputational Intelligence
EC
KE FS
NN
Powerful cooperative technologies were developed in the end of 20th Century
What Comes Next
System optimization based on human evaluation
Computer support system for creativity psychological and physical satisfaction
EC
KE FS
NN
+computer real human
Knowledge KANSEI
reasoning
expression
acquisition
associative memory
learning
intuition
preference
subjective evaluation
perception
cognition
etc etc
Two DifferentHuman Capabilities
explicitknowledge
experience
preference
computationalintelligence
optimization
auto-design
implicitknowledge
1st Generation
2nd Generation
3rd Generation
Importance of Human Factors withComputational Intelligence
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
What Comes Next
System optimization based on human evaluation
Computer support system for creativity psychological and physical satisfaction
EC
KE FS
NN
+computer real human
Knowledge KANSEI
reasoning
expression
acquisition
associative memory
learning
intuition
preference
subjective evaluation
perception
cognition
etc etc
Two DifferentHuman Capabilities
explicitknowledge
experience
preference
computationalintelligence
optimization
auto-design
implicitknowledge
1st Generation
2nd Generation
3rd Generation
Importance of Human Factors withComputational Intelligence
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Knowledge KANSEI
reasoning
expression
acquisition
associative memory
learning
intuition
preference
subjective evaluation
perception
cognition
etc etc
Two DifferentHuman Capabilities
explicitknowledge
experience
preference
computationalintelligence
optimization
auto-design
implicitknowledge
1st Generation
2nd Generation
3rd Generation
Importance of Human Factors withComputational Intelligence
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
explicitknowledge
experience
preference
computationalintelligence
optimization
auto-design
implicitknowledge
1st Generation
2nd Generation
3rd Generation
Importance of Human Factors withComputational Intelligence
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Humanized technology
ROBOTICSCONTROL
DATAMINING
SIGNALPROCESSING
natural environment human environment
acquired knowledge
physical measurement
qualifying the knowledge
human perception
locationspeedobstacleshellip
preferencesubjective evaluationemotionhellip
database miningIF hellip THEN hellipIF hellip THEN hellip
filter design SN ratio
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
1980 1990 2000
NN
FS
EC
NN+FS
GA+FS
NN+FS+GA
GPKoza (1980s)
NN-driven Fuzzy Reasoning (1988)
Karr et al (1989)
CI to be a competitor of humans
CI as human models
CI for human
one research direction is
Direction of Computational Intelligence
HumanizedComputational
Intelligence
Interactive Evolutionary Computation
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Part 2 What is IEC
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
What is IEC
Target System
evolutionarycomputation subjective evaluation
(1) fewer population size and fewer generation numbers(2) nevertheless satisfactory solutions can be obtained ie simple landscape(3) discrete fitness in n-evaluation levels and(4) relative fitness in general
Features
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
searching parameter space
fitne
ss
searching parameter space
fitne
ss
normal EC search interactive EC search
Searching spaces of IEC tasks are generally simple
because Any searching points that human operators cannotdistinguish are same for human
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
80s 90 91 92 93 94 95 96 97 98 99 00 total
graphic art amp CG animation 2 3 2 4 5 5 2 2 4 9 4 423-D CG lighting design 1 3 1 5music 1 3 3 1 1 3 5 17editorial design 1 1 2 4industrial design 2 2 1 5 4 2 4 9 29face image generation 1 1 1 2 1 4 5 1 16speech processing amp prosodic control 2 1 2 1 1 7hearing aids fitting 2 7 5 14virtual reality 1 1 2database retrieval 2 1 8 8 1 20knowledge acquisition amp data mining 5 3 3 1 4 16image processing 1 2 3control amp robotics 1 2 3 4 4 14internet 1 2 1 4food industry 1 1 2geophysics 1 2 3art education 2 2writing education 1 3 4games and therapy 1 1 1 3social system 1 1
discrete fitness value input method 5 2 7prediction of fitness values 1 2 1 8 3 1 16interface for dynamic tasks 1 1 3 5acceleration of EC convergence 1 1 3 1 7combination of IEC and non-IEC 1 2 3active intervention 1 3 2 6
total 2 0 5 5 8 11 23 28 22 48 57 43 252
Statistics of IEC Papers
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
IEC Research Takagi Lab
(1) 3-D CG lighting design support(2) montage image system(3) speech processing(4) image processing(5) hearing-aidcochlear implant fitting(6) virtual reality in robot control(7) media database retrieval(joint1) virtual aquarium(joint2) geoscientific simulation(joint3) 3-D CG modeling education(joint3) fireworks animation design(joint4) mental disease diagnosis
(1) input interface 11 discrete fitness value input method(2) display interface 21 prediction of users evaluation chars 22 display for time-sequential tasks(3) acceleration of EC convergence 31 approximation of EC landscape 32 IDEgravity IDEmoving vector(4) active user intervention to EC search 41 on-line knowledge embedding 42 Visualized IEC(5) IEC user model(6) New IEC frameworks
application-oriented interface research
IEC for Human Science(1) measuring human mind(2) finding unknown knowledge(3) IEC with physiological responces
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
IEC Research Categories
art educationwriting educationgames and therapysocial system
speech processinghearing aids fittingvirtual realitydatabase retrievaldata miningimage processingcontrol amp roboticsinternetfood industrygeophysics
graphic art amp CG animation3-D CG lighting designmusiceditorial designindustrial designface image generation
discrete fitness value input methodprediction of fitness valuesinterface for dynamic tasksacceleration of EC convergencecombination of IEC and non-IECactive interventionVisualized IEC
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Part 3 IEC Applications
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Hideyuki Takagi Interactive Evolutionary Computation Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275-1296 (Sept 2001)
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
System designed by Tatsuo UNEMI
Interactive GP for Graphic Art
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Breeding Forms on the Infinite Planeby W Latham (1992)
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Cheerful impressionGloomy impression Heroine movie star Wicked movie star
by HAND
by Interactive GA
IGA for 3D CG Lighting DesignBy K Aoki and H Takagi
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Result of Subjective and Statistic Tests
1 level of significance 5 level of significance
cheerful impression
IGA-based lighting design looks effective for beginners
heroine impression
villain impression
gloomy impression
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Expanding Lighting EnvironmentsThanks to Increasing LED Performanceand Decreasing LED Costs
Lighting environments expandsthanks to Increasing LED performanceand decreasing LED costs
Increasing the of LED lights makes lighting design difficult
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
IEC-based Lighting DesignIEC-based Lighting Design
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Education of Imagination and Creativity for 3D Shape
It is difficult and time consuming for computer beginners (artist non-technical student etc) to acquire 3D modeling skill
Apply IEC to support CG skill and focus on educating imagination and creativity
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
3D Model Example
blenddeform
P2
P3
P1
P5
P4
P5 P4 P3 P6
P1 P2
IEC changes shape parameters based on users imagination
Nishino et al (SMC1998 SMC1999)
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
IEC-base 3D Modeling Concept
designer
3D shape image to create
rateeach
shapevery goodgoodnot good
GeneticAlgorithm
create new 3D geometries by inheriting highly rated shapes and simulating natural evolutionary processes like crossover and mutation
display a set of new shapes
Nishino Takagi and Utsumiya (2000)
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Subjective Test
RealisticExamination
Creative Examinationferocious green pepper
Quality
Operability
Manual gtgt IEC
Manual gt IEC
Manual = IEC
Manual lt IEC
Only 48 (8 parameters times6 primitives) parameters are modified by GA
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Internet version of IEC 3-D Modeling Education System
car designers
students artists
IEC-based Modeling Interface
CG System for Car Body
Design
CG System
for Education
CG Systemfor Artwork
Creation
specify rates based on subjective preference
explore CG parameters by evolutionary computation
visualize 3D models by explored parameters
network
display generated models
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
IEC
Objective Fitness Function
fitness value
Evolution of Funny Animated Figuresby Jeffrey Ventrella
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
CG Fireworks Animation Made by IEC
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
type and position
of one of 32
foreheads
shape and position of
nose
shape and position of
mouth
shape and position of
chin
eyes and their
separationX X X X
7 bit 7 bit 7 bit 7 bit 7 bit
= 34 X 1010
faces
Montage Imageby C Caldwell and VS Johnston (1991)
target face montage face three days later 10 GA generations
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
hair
eyebroweyes
nose
mouth
face basement
Eyes look good
Then Id like to fix the eyes
eyehair
nose mouth
eyebrow
outline
by H Takagi and K Kishi (KES1999)
Montage Image
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
IGA for Music
GenJam
piano drum
code progression
phrase population
measure population
MIDI sequence
bass
goodbad
solo
by J A Biles (1994)GenJam Jazz Solo Generator
Rhythm Generator by D Horowithz (1994)
Sonomorphs Melody Generator by G L Nelson (1993)
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Industrial Designby H Furuta
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
HTML Design
ECfont of letters colors of letters and background and etc
Looks nice
by N Monmarche et al
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Changing Colors and Fonts with IEC
Making Layoutwith IEC
Stability 10Power 05etc
Vigor 10 etc
Completed Poster
By T Obata and M Hagiwara
Color Poster Design
Colors and fonts are VIGOR and layout is STABILITY
Colors and fonts are CLEAN and layout is ELEGANT
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Q1 Q2R1
R2
EC
Visitors who click the banner
by R Gatarsk
Evolutionary Banner
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Why IEC-based Signal Processing
There are many cases that SP users are not SP experts but need to design SP filters
image filter hearing aid
Solution is auditory-based SP and visual-based SP without any SP knowledge
IEC realizes this approach
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Recovering Distorted Speech
Distorted SpeechOriginal Speech
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10
Pow
er (
dB)
Frequency (kHz)
Distorted Speech
A B
C
three operators
Interactive GAa1 a2 a4 a6a0 a7a5a3
8 bits
H(z) = Σa z kk = 0
-k7
Watanabe and Takagi (SMC2005)
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Experimental Result
-- Method of Successive Categories --
distance scale
Ideal Situation GA filter designwhen ideal speech is given
40
-20 -10 00 10
A20
C20
A40
C40
B20
B10
B
C10
A10
20
distorted speech
poora little poor neutral a little
good good10{
Sheffersquos method of paired comparison showed that this difference is significant (plt001)
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Experimental Result
Sheffersquos Method of Paired Comparisons
-10 -05 00 05 10
A
B
C 10
20 40 10
20
401020
40
poorer better
distorted speech
10th generation of recovering speech
20th generation of recovering speech
40th generation of recovering speech
10
20
40
A
B
C
vs vsvsvs vsvs40 1020 4010 20 401020operators
combinations
difference is significant (plt001)
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
How Distorted Speech was Recovered
Formant area was mainly recovered
(i) A
A10
A20
A40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(ii) B
B10
B20
B40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
20
(iii) C
20
C10
C20
C40
0 1 2 3 4 5 6 7 8 9 10Frequency (kHz)
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
amplitude
pitch
tempo
speech synthesizeroriginal
speechspeech analysis
spectral information
ECsubjective evaluation
prosodic control
Voice Conversion by Interactive ECby Y Sato et al
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
Example of Prosodic Modificationby Y Satoh et al
Original Voice Modified by IEC
Text-to-Speech Intelligible voice
Text-to-Speech male-like voice
Intelligible voice
real voice child-like voice
real voice
By T Morita S Iba and M Ishizuka
IEC for Agentrsquos Voice Design
EC
eval
uatio
n
Proso
dy co
ntro
l
I cannot hear both
Hz100 1K 10K
0
20
40
60
80
100
120
dB
Hearing Compensation is Difficult
cognitive levelperceptual level
sensory level
preference
final hearing
input sounds
measurable in part
target goal
Ideal Goal is Far
hearing aid
Conventional
EC
hearing aid
Proposed
Tuning System based on How User hears
Ohsaki and Takagi (1998-2007)
Evaluation
evaluation IEC Fitting vs audiologist fitting
monosyllable articulation
sound quality
fitting time IEC lt audiologist
One ndash Two Weeks Later (5 subjects)
Six Months Later (4 subjects)
evaluation IEC Fitting vs audiologist fitting
sound quality
APHAB
Visualized IEC Fitting on a PDA
Visualized EC parameters in an n-D EC landscape are mapped on a 2-D space for visualization
IEC Fitting IEC-based hearing aid fitting
Hayashida and Takagi (IECON2000)
Image Enhancement using
Interactive GPby R Poli and S Cagnoini
Image enhancement filter evolves according to how processed images look well
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
I cannot hear both
Hz100 1K 10K
0
20
40
60
80
100
120
dB
Hearing Compensation is Difficult
cognitive levelperceptual level
sensory level
preference
final hearing
input sounds
measurable in part
target goal
Ideal Goal is Far
hearing aid
Conventional
EC
hearing aid
Proposed
Tuning System based on How User hears
Ohsaki and Takagi (1998-2007)
Evaluation
evaluation IEC Fitting vs audiologist fitting
monosyllable articulation
sound quality
fitting time IEC lt audiologist
One ndash Two Weeks Later (5 subjects)
Six Months Later (4 subjects)
evaluation IEC Fitting vs audiologist fitting
sound quality
APHAB
Visualized IEC Fitting on a PDA
Visualized EC parameters in an n-D EC landscape are mapped on a 2-D space for visualization
IEC Fitting IEC-based hearing aid fitting
Hayashida and Takagi (IECON2000)
Image Enhancement using
Interactive GPby R Poli and S Cagnoini
Image enhancement filter evolves according to how processed images look well
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
cognitive levelperceptual level
sensory level
preference
final hearing
input sounds
measurable in part
target goal
Ideal Goal is Far
hearing aid
Conventional
EC
hearing aid
Proposed
Tuning System based on How User hears
Ohsaki and Takagi (1998-2007)
Evaluation
evaluation IEC Fitting vs audiologist fitting
monosyllable articulation
sound quality
fitting time IEC lt audiologist
One ndash Two Weeks Later (5 subjects)
Six Months Later (4 subjects)
evaluation IEC Fitting vs audiologist fitting
sound quality
APHAB
Visualized IEC Fitting on a PDA
Visualized EC parameters in an n-D EC landscape are mapped on a 2-D space for visualization
IEC Fitting IEC-based hearing aid fitting
Hayashida and Takagi (IECON2000)
Image Enhancement using
Interactive GPby R Poli and S Cagnoini
Image enhancement filter evolves according to how processed images look well
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
hearing aid
Conventional
EC
hearing aid
Proposed
Tuning System based on How User hears
Ohsaki and Takagi (1998-2007)
Evaluation
evaluation IEC Fitting vs audiologist fitting
monosyllable articulation
sound quality
fitting time IEC lt audiologist
One ndash Two Weeks Later (5 subjects)
Six Months Later (4 subjects)
evaluation IEC Fitting vs audiologist fitting
sound quality
APHAB
Visualized IEC Fitting on a PDA
Visualized EC parameters in an n-D EC landscape are mapped on a 2-D space for visualization
IEC Fitting IEC-based hearing aid fitting
Hayashida and Takagi (IECON2000)
Image Enhancement using
Interactive GPby R Poli and S Cagnoini
Image enhancement filter evolves according to how processed images look well
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Evaluation
evaluation IEC Fitting vs audiologist fitting
monosyllable articulation
sound quality
fitting time IEC lt audiologist
One ndash Two Weeks Later (5 subjects)
Six Months Later (4 subjects)
evaluation IEC Fitting vs audiologist fitting
sound quality
APHAB
Visualized IEC Fitting on a PDA
Visualized EC parameters in an n-D EC landscape are mapped on a 2-D space for visualization
IEC Fitting IEC-based hearing aid fitting
Hayashida and Takagi (IECON2000)
Image Enhancement using
Interactive GPby R Poli and S Cagnoini
Image enhancement filter evolves according to how processed images look well
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Visualized IEC Fitting on a PDA
Visualized EC parameters in an n-D EC landscape are mapped on a 2-D space for visualization
IEC Fitting IEC-based hearing aid fitting
Hayashida and Takagi (IECON2000)
Image Enhancement using
Interactive GPby R Poli and S Cagnoini
Image enhancement filter evolves according to how processed images look well
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Image Enhancement using
Interactive GPby R Poli and S Cagnoini
Image enhancement filter evolves according to how processed images look well
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
+ =diastole of patient A systole of patient A
(1)
(2)
(3)
0832041 g1g2 2g1 2g2 g1
2g2 g2 min( 0277346 g1 g2max(g1g2 g1))
0554695 g1g2 g12g2 g2 g1 g2
g1g2 g22 2g2 2g1g2g1
2 064861 g2 g1
g1g2 g22 2g2 3g1 g1g2 0854702 min(2g2 2g1 g1(g2 g1))
(1)
(2)
(3)
(4)
(4)
Echo-Cardiographic Image
by R Poli and S Cagnoni (1997)
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Non-linear Density Transformation by Using of GP
linear density transformation
non-linear density transformation by using GP
exinvert g(ij)
output
output
input
inputg(ij)= G ( f ( i j ) )
generated by GP
ou
tpu
t d
en
sity
input densityo n f(ij)
n
ou
tpu
t d
en
sity
input densityo n f(ij)
n
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
GP-Based Image Filter by Using Neighborhood Pixels Information
f(i j) g(i j)
input pixels output pixels
operator G
g(ij) = G(f)
x
y
x
y
GP generates G( )
zoom in
G=max[abs(f(ij-1))max-041079+sin(min(f(i+1j-1)-003))-f(i+1j+1)]
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Experiment of IGP-Based Edge Detection
input image
Laplacian filter 2nd generation
conventional math-based filter IEC-based filter
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
high-pass filter 6th generation
Experiment of IGP-Based High Pass Filter Design
input image
conventional math-based filter IEC-based filter
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
x-axis pass filter
input image
Prewitt 4th generation
conventional math-based filter IEC-based filter
IGP-Based X-component Extraction
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
6th generationPrewitt
input image
y-axis pass filter
conventional math-based filter IEC-based filter
IGP-Based Y-component Extraction
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC-based Color Filter Design
input image
examples of coloring
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Image Enhancement Filter Design by a Medical Doctor
original ultrasonic image of a lymph node
Enhanced Image after 12 IEC generations
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Consumer Robots BusinessHas Been Launched
Home robots will come soon
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC
Parameters
fearful
secure
Human Friendly Trajectory Control of a Robot Arm
by N Kubota K Watanabe and F Kojim
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
by H H Lund O Miglimo L Pagliarini A Billard and A Ijspeert
yj = Σ wijxi + w0j
n
i=1
infrared sensor 1
infrared sensor 2
mechanical switch 1
mechanical switch 2
mechanical switch 3
mechanical switch 4
1 (offset)
6 c
on
tro
l ou
tpu
ts
NN Controller for LEGO Jeep-Robot
1 Children want to make a robot avoiding obstacles
2 Children cannot make a program of its controller but can choose better robot moving
3 Lets evolve the robot controller according to the childrens choice
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Interactive EC for Virtual Reality
Arm wrestling human vs robot
controller
fuzzy control rules for VR
IECsubjective evaluation for VR
modifying rules
Kamohara Takagi and Takeda (SMC2007)
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC for Dental Education in VR
How to realize the reality of VR force feedback
VR dental education software
httpjks-folksstanfordedubonesim
httpelianealhadeffblogspotcom200905dental-serious-games-with-3d-touchhtml
httpwww1800dentistcomdentist-love-blogblogdefaultaspxuseridpost=3
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Media DB Retrieval amp Media Converter
NN
psychological
GA search
active
complexity
space factor
pitch
duration
variance
featurespace
space
My impression is hellip
by H Takagi et al
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC-based Image DB Retrievalby S-B- Cho et al
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
2300 questionnaire for oral care goods
16 image words 2300
16 goods attributes
inductive learning tool
imag
e w
ords
vs
goo
d at
trib
utes
re
latio
n tr
ee
GA
relation tree 1
relation tree 2
relation tree 3
relation tree 4
relation tree n
GA creates variation of image-attribute relation trees
Human chooses simpler and better relations of image-attribute relations
Interactive EC for Data Mining
by T Terano Y Ishino et al
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Visual Data-miningThrough 2-D Mapping by GP
X = (x1-u) (873 x12) Y = (467 x6) (252 + 810)for example
by Venturini
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Intrusion Pattern Generationsfor Intrusion Detection System
IDS
How to evaluate an intrusion Detection system
Ja-Min KOO and Sung-Bae CHO (2004)
virus
illegal attack
computer virus
How to generate several new types of intrusion patterns for evaluation
virus
virus
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
J Budynek E Bonabeau and B Shargel (2005)
script individualsgene pool
Generating Intrusion Patternsfor Intrusion Detection Systems
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
crossover
mutation
deletion
J Budynek E Bonabeau and B Shargel (2005)
Generating Intrusion Patternsfor Intrusion Detection Systems
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Intrusion Pattern Generationsfor Intrusion Detection System
IntrusionDetectionSystem
Ja-Min KOO and Sung-Bae CHO (2004)
generatedintrusion patterns
+
GA
detection performancecode evaluation
(A)send ldquolynx zezeonelove64comtemp6htmrrdquoexpect ldquoDownloadrdquosend ldquodrrdquosend ldquo010010010crrdquosleep 1send ldquoyrdquosend ldquoqyrdquo
(B)send ldquovi tempcrrdquosend ldquo1$ s(064)agnrdquosend ldquo1$ s(017)bgnrdquosend ldquo1$ s(003)cgnrdquosend ldquo1$ s(041)dgnrdquosend ldquo1$ s(081)egnrdquosend ldquo1$ s(027)fgnrdquosend ldquo1$ s(055)ggnrdquosend ldquo1$ s(060)hgnrdquosend ldquo1$ s(177)ignrdquosend ldquo1$ s(361)jgnrdquosend ldquo1$ s(011)kgnrdquosend ldquo1$ s(452)lgnrdquosend ldquo1$ s(913)mgnrdquosend ldquo1$ s(784)ngnrdquosend ldquoqnrdquosleep 1
(C)send ldquovi Makefilerrdquosend ldquo1$ shelloctempcgrrdquosend ldquoqnrdquosend ldquomakerrdquosend ldquohellorrdquo
(D)send ldquocp etcshadow etcshaddowrrdquosend ldquochmod o+r etcshadowrrdquo
(E)send ldquorm -rf etcshadownrdquosend ldquocp etcshaddow etcshadowrrdquo
73 of generated executable intrusion patterns were different
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 5 IEC for Human Science
CONTENTS
1 Artistic ApplicationsCG lighting design CG animation montage music industrial design
2 Engineering Applicationssignal processing robotics VR media DB retrieval data mining intrusion detection
3 Othersgeo-science coffee blending game amp education
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
a priori geological knowledge subjective evaluation
Visualization of geological history based on L
Moresis forward code observed current condition
estimated initial condition of
geological features
Evolutionary Computation
geoscientist
mod
el
par
ame
ters
Geological ModelingBased on Interactive EC
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Difficulty
bull many variablesndash strength depth stress etc
bull no numerical targetndash Numerical similarity may not mean qualitative similarity such as wrong
fault inclination wrong depth extentndash Help of geologists is necessary
bull computational optimisation with human judgementndash Computational optimisation methods to search material parameters are
required besides geologistrsquos judgement
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
sketch drawn bya geologist
geologicalforwardmodel
EC evaluation based ongeological knowledge and experience
mat
eria
lpa
ram
eter
s
IEC-based Modelling of Extension of the Earthrsquos Crust
Wijns Moresi Boschetti and Takagi (SMC2001)
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC-based Modelling of Subduction of Oceanic Crust
animations
geologicalforwardmodel
ECevaluation based on geological
knowledge and experience
mat
eria
lpa
ram
eter
s
Strength of continentaloceanic crust Convergence rate Depth of sedimentary wedge
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Mocha Kilimanjaro
Colombia
Target Brended Coffee
IEC
Fitness Value
Mixture Ratio
Evolving Blends of Coffeeby M Hardy
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
I go to school on foot
Today I took a bus and
went to school
I enjoyed school swimming
Satomi is a girl who loves athletic meet
She commutes by bus every day
When she arrived at her school
She found there was an athletic meet
Simulated Breeding forComposition Support System
by K Kuriyama T Terano and M Numao
Two sequences of four scenes are chosen as parents
Two better sequencesare chosen as parents
Composition and comparison with similar composition
INITIAL DISPLAY EVOLVED SEQUENCES FINAL DISPLAY
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 4 IEC research for Reducing IEC User Fatigue
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Approaches for Reducing IEC User Fatigue
(1) input interface
11 discrete fitness value input method
(2) display interface 21 prediction of users evaluation chars
22 display for time-sequential tasks
(3) acceleration of EC convergence 31 approximation of EC landscape
32 IDEgravity IDEmoving vector
(4) active user intervention to EC search 41 on-line knowledge embedding
42 Visualized IEC
(5) IEC user model
(6) New IEC frameworks
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 5 IEC for Human Science
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Psychotherapy Diagnostics
IEC-based CGLighting Design
mental patient(schizophrenics)
sad happy
sad happy
emotion range of mentally healthy persons
emotion range of the patient
psychotherapistpsychiatrist
Umm stillhis range is narrower
Takagi Takahashi and Aoki (SMC2004)
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
PM
PK
NH
NS
NY
NK
happy impression
NN
PT
happy scale
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
NY
NK
NN
NS
NH
PM
PK
PT
sad impression
sad scale
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
DiscussionDiscussionobtain the order scale of happy - sad expression range from psychological scales of happy and sad obtained through a subjective test
NS
NN
NK
PM
NY
PT
NH
PK
1st 5th
2nd
3rd
4th
6th
7th
8th
sad happy sad happy
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Cochlear Implants FittingP Legrand C Bourgeois-Republique et al
conventionalfitting
E1 E4 E5E2 E3 E6 E7 E9E8 E13E14E15
20
15
10
5
electric channels
inte
nsity
IEC fitting
eval
uatio
nfitting
45
915 (IEC fitting)
485 (practitionerrsquos result)
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC Breaked Common Sense
0
5
10
15
20
25
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Setting all electrodes but E1E7 E9 under the liminary value gives 821
0
2
4
6
8
10
12
14
16
18
20
22
1 2 3 4 5 6 7 8 9 13 14 15
Adding electrodes E4 and E15 gives 588 More electrodes better audition
5678910111213141516171819202122
E 1 E 2 E 3 E 4 E 5 E 6 E 7 E 8 E 9 E 13 E 14 E 15
Only 3 electrodes (E1 E7 E9) have 3 a significant C-T rangebut the best 915
Slide made by Prof Pierre Collet
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Obtained Knowledge from Analysis of IEC Fitting
Loudness charact tuned by speech (IEC fitting)ne
fitting using speech1
wwo noise
fitting using speech2
wwo noise
fitting using speechN
wwo noise
fitting using music wwo
noisene
differences are small
(1)
(2)
The best hearing characteristics of human seems to depend on sounds
EC
hearing aid
Loudness charact tuned by noise or pure tones (conventional fitting)
M Ohsaki and H Takagi (1999)
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 5 IEC for Human Science
CONTENTS
1 measuring dynamic range of happy-sad scale
2 finding unknown psychophysiological facts
3 extended IEC IEC with physiological responses
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
EmotionEnhancement
Filter
Example Application
physiologicalresponses
control
physicalfeatures
Emotion Controlwith 3 Steps
IEC with PhysiologyIEC with PhysiologyH takagi (2004)
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Step 3 Conventional IEC Based on Psychological Feedback
Step 3 Conventional IEC Based on Psychological Feedback
EC
imagefilters
Conventionalsubjective evaluation
IEC = system optimization technique based on human subjective feedback
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Step 3 Extended IEC Based onPhysiological Feedback
Step 3 Extended IEC Based onPhysiological Feedback
EC
imagefilters
Proposal-
physiologicalmeasurement
given targetphysiological
conditions
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 6 New IEC framework
1 Humanized Computational Intelligence2 What is IEC3 IEC Applications4 IEC research for Reducing IEC User Fatigue 5 IEC for Human Science6 New IEC framework
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Multi-Objective Optimizationwith Subjective Object(s)
Im looking for an apartment that is(1) wider (2) less expensive and (3) closer to a station
wid
th
tenant width
dist
ance
to
st
tena
nt
distance to st
Then how to find an apartment (4) with great view (5) in a quite area
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Multi-Objective Optimization MEMS Design
joint research with UC Berkeley
MEMS (Micro Electronic Mechanical Systems) for Sensors Robotics Communications Biotechnology Energy Generation
bull Multi-objective optimization for given specification receiving frequency strength and othersbull We want to use domain knowledge for circuit
design
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Raffi Kamalian et al (2004)
bull EMO four objects1 Resonant freq = 100kHz2 K_vertical = 1 Nm3 K_lateral = 100 Nm4 Area = minimized
bull IEC embedding domain knowledge
for MEMS circuit design
IEC + EMO for MEMS DesignIEC + EMO for MEMS Design
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC+EMO EMOUser Expert of 5s of 5s sign
1 Y 7 9 -12 Y 12 6 13 Y 7 3 14 N 6 2 15 Y 4 4 06 Y 11 9 17 N 8 7 18 Y 1 0 19 N 6 3 110 N 12 7 111 N 9 2 1
Subjective Test + Statistic Testfor EMO+IEC vs EMO
Subjective Test + Statistic Testfor EMO+IEC vs EMO
EMO+IEC gt EMO (99 significant by Wilcoxon matched-pairs signed-ranks test )
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IEC + EMO for Architectural DesignIEC + EMO for Architectural Design
httpwwwmlitgojpkishakisha0207071220_html
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
106
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Algorithm for Generating LayoutAlgorithm for Generating Layout
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda
Corridor
W
K
Wall
L
B
Verenda 107
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
IEC +EMO for Room Layout-- Algorithm for Generating Layout --
Makoto Inoue et al (2008)
wal
l
wal
l
wal
lw
all
wal
l
wal
l
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Corridor
L B
W
Wall
K
Verenda
Corridor
W
B
L Wall
K
Verenda
Corridor
W Wall
B K
Verenda
Corridor
L W Wall
B K
Verenda
Corridor
L W
B Wall
Verenda
Corridor
W
K
B Wall
L
Verenda
Layout Algorithm
objective 1
obje
ctiv
e 2
EC
EMO IEC
IEC +EMO for Room LayoutIEC +EMO for Room Layout
1 room sizes2 room shapes3 moving line to any rooms without
passing through other private rooms4 window sizes requested by
architectural laws5 total water pile length6 total wall length
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
1
32
Particle Swarm OptimizationParticle Swarm Optimization
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
GA PSOF1 F2 F3 F4 F5 F6 ≒F7 F8
Comparison of Performance
GA vs POS and IGA vs IPOSGA vs POS and IGA vs IPOS( faster ≒ equivalent )
com
plex
ity
high
low
IGA IPSO
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
54321
5-levels
10-levels
100-levels
1000-levels
quan
tizat
ion
nois
ein
fitn
ess
Tolerance of PSO forQuantization Noise in Fitness
Tolerance of PSO forQuantization Noise in Fitness
fewer
more
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
0
1
2
3
4
5
6
7
1 11 21 31 41
5段階10段階100段階1000段階
Interactive PSO
levels
levels
levels
levels
levels
levels
levels
levels
Experimental Result for F1Experimental Result for F1
Interactive GA
F2 ~ F8 has similar tendency
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IGA
IPSO
IPSO with proposed methods
generations
fitne
ssAfter Implementing Four Methods for
Increasing Noise ToleranceNakano and Takagi (CEC2009)
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Part 6 New IEC framework
CONTENTS
1 IEC + EMO
2 Interactive PSO and interactive DE
3 paired comparison-based IEC
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Pair Comparison Tournament IECPair Comparison Tournament IEC
Ordinary IEC
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Paired Comparison-basedInteractive Differential Evolution
Paired Comparison-basedInteractive Differential Evolution
indi
vidu
als
targetvector
trialvector
targetvector
parametervector 1
parametervector 2
basevector
mutantvector
F
Takagi and Mitsuyasu (patent)Takagi and Palles(CEC2009)
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
IGA vs Paired Comparison-based IDE
GA
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
HA 1
HA 2HA 1DE
ON ON
HA 5 HA 6
HA 9 HA 10 HA 11 HA 12
HA 13 HA 14 HA 15 HA 16
HA 3
HA 8
HA 2
HA 7
HA 4
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
CONCLUSION
bull One of IEC research directions is Humanized Computational Intelligence and IEC is one of tools for this direction
bull IEC has high applicability in wide variety of areas
bull IEC research includes research reducing IEC user fatigue and research for human science besides major IEC research IEC applications
bull We also introduced recent new IEC research on extending IEC algorithms and frameworks
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-
Further Information
bull IEC tutorial amp survey paperndash Hideyuki Takagi Interactive Evolutionary Computation
Fusion of the Capacities of EC Optimization and Human Evaluationrdquo Proceedings of the IEEE vol89 no9 pp1275--1296 (Sept 2001)
ndash Its draft paper is downloadable from the below URL
bull IEC slides are downloadable from the below URL
bull Personal Contactndash takagidesignkyushu-uacjpndash httpwwwdesignkyushu-uacjp~takagi
- Interactive Evolutionary Computation
- Part 1 Humanized Computational Intelligence
- Computational Intelligence in the 20th Century
- Cooperation of Computational Intelligence
- What Comes Next
- Two Different Human Capabilities
- Slide 7
- Slide 8
- Direction of Computational Intelligence
- Part 2 What is IEC
- What is IEC
- Searching spaces of IEC tasks are generally simple because
- Statistics of IEC Papers
- IEC Research Takagi Lab
- IEC Research Categories
- Part 3 IEC Applications
- Part 5 IEC for Human Science CONTENTS
- Interactive GP for Graphic Art
- Breeding Forms on the Infinite Plane
- IGA for 3D CG Lighting Design
- Result of Subjective and Statistic Tests
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Internet version of IEC 3-D Modeling Education System
- Evolution of Funny Animated Figures
- CG Fireworks Animation Made by IEC
- Montage Image
- Slide 32
- IGA for Music
- Industrial Design
- HTML Design
- Color Poster Design
- Evolutionary Banner
- Part 5 IEC for Human Science CONTENTS (2)
- Why IEC-based Signal Processing
- Recovering Distorted Speech
- Experimental Result
- Experimental Result (2)
- How Distorted Speech was Recovered
- Voice Conversion by Interactive EC
- Example of Prosodic Modification
- IEC for Agentrsquos Voice Design
- Hearing Compensation is Difficult
- Ideal Goal is Far
- Tuning System based on How User hears
- Evaluation
- Slide 51
- Image Enhancement using Interactive GP
- Echo-Cardiographic Image
- Non-linear Density Transformation by Using of GP
- GP-Based Image Filter by Using Neighborhood Pixels Information
- Experiment of IGP-Based Edge Detection
- Experiment of IGP-Based High Pass Filter Design
- Slide 58
- Slide 59
- IEC-based Color Filter Design
- Slide 61
- Consumer Robots Business Has Been Launched
- Human Friendly Trajectory Control of a Robot Arm
- NN Controller for LEGO Jeep-Robot
- Slide 65
- Interactive EC for Virtual Reality
- IEC for Dental Education in VR
- Media DB Retrieval amp Media Converter
- IEC-based Image DB Retrieval
- Interactive EC for Data Mining
- Visual Data-mining Through 2-D Mapping by GP
- Intrusion Pattern Generations for Intrusion Detection System
- Slide 73
- Slide 74
- Intrusion Pattern Generations for Intrusion Detection System (2)
- Part 5 IEC for Human Science CONTENTS (3)
- Geological Modeling Based on Interactive EC
- Difficulty
- IEC-based Modelling of Extension of the Earthrsquos Crust
- IEC-based Modelling of Subduction of Oceanic Crust
- Evolving Blends of Coffee
- Simulated Breeding for Composition Support System
- Part 4 IEC research for Reducing IEC User Fatigue
- Approaches for Reducing IEC User Fatigue
- Part 5 IEC for Human Science
- Part 5 IEC for Human Science CONTENTS (4)
- Psychotherapy Diagnostics
- Slide 88
- Slide 89
- Slide 90
- Part 5 IEC for Human Science CONTENTS (5)
- Slide 92
- IEC Breaked Common Sense
- Obtained Knowledge from Analysis of IEC Fitting
- Part 5 IEC for Human Science CONTENTS (6)
- Slide 96
- Slide 97
- Slide 98
- Part 6 New IEC framework
- Part 6 New IEC framework CONTENTS
- Slide 101
- Slide 102
- Slide 103
- Slide 104
- Slide 105
- Slide 106
- Slide 107
- Slide 108
- Part 6 New IEC framework CONTENTS (2)
- Slide 110
- Slide 111
- Slide 112
- Slide 113
- After Implementing Four Methods for Increasing Noise Tolerance
- Part 6 New IEC framework CONTENTS (3)
- Slide 116
- Slide 117
- IGA vs Paired Comparison-based IDE
- CONCLUSION
- Further Information
-