Short-term memoryLong-term memory

Structure of rule based expert system

Production Rules Facts

Reasoning

Conclusion

Inference engine match-fire procedure

Knowledge base

Database

Fact: A is x Fact: B is y

Rule: If A is x then B is y

Match Fire

Probability of A

A

Probability of B

B

Joint probability of A and B

A A B B

Boolean logic examples

• Quick cars run a14.1 second quarter mile or better.

• People with hairy backs have 2500 or more hairs on their backs.

• Motors that run really hot are 245 degrees or more.

SlowQuick

HairyBald

14.1 sec

2500 hairs

HotCold

245 degrees

14.1 seconds

2003 Camaro SS

1969 Camaro Z28 (14.2)

2006 Nissan Maxima (15.2)

1984 VW Diesel Rabbit (19.8)2007 Corvette Z06

(13.05)

1966 427 Cobra (11.5)

1988 Pro-stock Firebird(7.88)

Crisp “Tall Men”

Crisp 'Tall Men'

0

0.2

0.4

0.6

0.8

1

1.2

150 160 170 180 190 200 210

height (cm)

deg

ree

of

mem

ber

sh

ip

Crisp

Fuzzy “Tall Men”

Fuzzy 'Tall Men'

0

0.2

0.4

0.6

0.8

1

1.2

150 160 170 180 190 200 210

height

deg

ree

of

mem

ber

ship

Fuzzy

Fuzzy ‘short, average, tall men’

Fuzzy set, short, average, tall

0

0.2

0.4

0.6

0.8

1

1.2

155 160 165 170 175 180 185 190 195

height

mem

ber

ship

tall

short

average

Crisp ‘short, average, tall men’

0

0.2

0.4

0.6

0.8

1

1.2

155 160 165 170 175 180 185 190 195

tall

short

average

Start

Generate Population of Chromosomes of size N: x1, x2…xn

Calculate fitness of each Chromosome: f(x1), f(x2)…f(xn)

Terminate?

Select pair of chromosomes for mating

Crossover: p(c)

Mutation: p(m)

Add new chromosome to population

Is new Population complete?

Replace current population with new population

StopYes

No

Yes

No

Start

Generate Population of Parameters of size N: x(1) to x(n)

Calculate solution associated with parents: X = f(x(1)….x(n))

Replace X with X’

Stop

No

Get number of generations

Create a population of offspring parameters:x’(1) = x(1) + µx’(2) = x(2) + µ ; µ is a random peturbation

Calculate solution associated with offspring: X’ = f(x’(1)….x’(n))

Is X’ better than X ?

Number of generations complete?

No

yes

yes

Start

Generate Population of S-expressions of size N

Calculate fitness of each S-expression

Stop

Terminate?

No

yes

yesRank S-expressions by fitness

Transfer best S-expression to new population

Select genericoperator

Select one S-expression Select pair of S-expressions Select one S-expression

clone mutatecrossover

Breed

New populationComplete?

Replace old population with new population

Simple diagram of a Neuron

Soma

Synapse

AxonDendrites

Som

a

Soma

Dot product

x(1)

x(2)

x(..)

x(..)

x(n)

w(1)

w(2)

w(..)w(..)

w(n)

yTransferfunction