Basics of Neuromorphic Design Sylvain Saïghi IMS Lab – University of Bordeaux.
-
Upload
sherman-lawson -
Category
Documents
-
view
214 -
download
0
Transcript of Basics of Neuromorphic Design Sylvain Saïghi IMS Lab – University of Bordeaux.
Basics of Neuromorphic Design
Sylvain Saïghi
IMS Lab – University of Bordeaux
2
Outline
Integrated Circuit Design
Models for neuromorphic implementation
Main types of neuromorphic systems
GDR BioComp – Colloque 2015 – St Paul de Vence
3
How to design an Integrated Circuit? IC technology providers for prototyping and low volume
production
Circuits Multi-Projets http://cmp.imag.fr/
Europractice http://www.europractice-ic.com/
Same kind of providers in North America
GDR BioComp – Colloque 2015 – St Paul de Vence
4
How to design an Integrated Circuit? Schedule
GDR BioComp – Colloque 2015 – St Paul de Vence
5
How to design an Integrated Circuit? Schedule and Prices…
GDR BioComp – Colloque 2015 – St Paul de Vence
6
How to design an Integrated Circuit? Second step: the schematic
Design
GDR BioComp – Colloque 2015 – St Paul de Vence
7
How to design an Integrated Circuit? Second step: the schematic
Design
Simulation
Monte Carlo to
check the reliability
against the parameters
mismatch
GDR BioComp – Colloque 2015 – St Paul de Vence
8
How to design an Integrated Circuit? Third step: the layout
Mask drawings are sent
to the IC provider
The provider gathers all
chips from different
designers
GDR BioComp – Colloque 2015 – St Paul de Vence
9
How to design an Integrated Circuit? Third step: the layout
Mask drawings are sent
to the IC provider
The provider gathers all
chips from different
designers
You will receive 25 samples
3 or 4 months later
Total duration about 12-18 months
GDR BioComp – Colloque 2015 – St Paul de Vence
10
Another solution for digital design VHDL: Very High-Level Design Language
Descriptive language Behavioral simulation
Choice of target• IC (Founder library)• FPGA (Field
Programmable
Gate Array) Physical simulations
GDR BioComp – Colloque 2015 – St Paul de Vence
11
S o m a
A x o n
N o d e o f R an v ie r
M y e lin sh ea th
P o st-sy n ap ticce ll
P re -sy n ap ticce ll
S y n ap tic te rm in a lsA neuron for neuromorphic designers
GDR BioComp – Colloque 2015 – St Paul de Vence
A few figures Soma diameter ~ 4 to 100 mm Axonal time propagation ~ 1 to 30ms Membrane voltage from ~ -120mV to 50mV Ionic currents ~ 10 nA
12
A neuron for neuromorphic designers
GDR BioComp – Colloque 2015 – St Paul de Vence
Spiking Neuron → time dependent
In tegrator Thresholddetector
Spikegenerator
SYNAPTICCURRENTS
13
Hebbian rule: Spike Timing Dependent Plasticity (STDP)
Learning in Neuromorphic Systems
Nj
Ni
Presynaptic neuron
Postsynaptic neuron
Wji
Causality
Increasing of Synaptic Weight
(potentiation LTP)
No causality
Decreasing of Synaptic Weight (depression LTD)
14
Threshold = 2
1/3
1/3
1/3
1/3
1/3
1/3
15
2/3
Threshold = 2
2/3
2/3
2/3
2/3
2/3
16
1
Threshold = 2
1
1
1/3
1/3
1/3
17
1
Threshold = 2
1
2/3
0
0
0
18
1
Threshold = 2
1
0
0
0
1/3
19
1
Threshold = 2
1
0
0
0
0
20
First neuromorphic design Carver Mead – Caltech
The term neuromorphic was coined by Carver Mead, in the late 1980s to describe Very-Large-Scale Integration (VLSI) systems containing electronic circuits that emulate the bio-physics of neural systems using the physics of silicon.
GDR BioComp – Colloque 2015 – St Paul de Vence
1989
21
2 main types of neuromorphic design
GDR BioComp – Colloque 2015 – St Paul de Vence
DigitalAnalog
Neuromimetic Rebuild, understand the life
Biological Real Time (99% of case)
Could be connected to life cells
Bio-Inspired Engineering issues
Event based computation
?
?
?
?
22
Neuromimetic designNeuroGrid (Standford)
Simulate the brain
in real-time Specifications:
sixteen Neurocores
each of which has
256 x 256 silicon neurons in an 11.9 mm x 13.9 mm chip Analog core and Digital communication bus
GDR BioComp – Colloque 2015 – St Paul de Vence
“Neurogrid simulates a million neurons connectedby billions of synapses in real-time, rivaling a supercomputer while consuming a 100,000 times less energy—five watts instead of a megawatt!”
23
Neuromimetic designSpiNNaker (Manchester University) – Human Brain Project
Simulate the brain in real time
Specifications a million-core computing
engine massively-parallel fully digital design
GDR BioComp – Colloque 2015 – St Paul de Vence
24
Neuromimetic designHeidelberg’s system (Heidelberg University) – Human Brain Project
Simulate the brain 104 faster than real-time
Specifications Based around wafer-scale VLSI Analog core and Digital
communication bus Each 20-cm-diameter silicon wafer
contains 384 chips, each of which
implements 128,000 synapses and up to 512 spiking neurons.
Total of around 200,000 neurons and 49 million synapses per wafer.
GDR BioComp – Colloque 2015 – St Paul de Vence
25
Bio-Inspired design TrueNorth – IBM
Video treatment
Specifications Fully digital
GDR BioComp – Colloque 2015 – St Paul de Vence
26
Bio-Inspired design TrueNorth – IBM
TrueNorth video
GDR BioComp – Colloque 2015 – St Paul de Vence
27
Bio-Inspired design Zeroth – Qualcomm
Neural Processing Units
which learn your abs
Specifications Fully digital
GDR BioComp – Colloque 2015 – St Paul de Vence
28GDR BioComp – Colloque 2015 – St Paul de Vence
Qualcomm video
29
Other neuromorphic research groups(non-exhaustive)
Europe• Institute of Neuroinformatics (Zurich)• Instituto Microelectronica (Sevilla)• University of Bordeaux• …
North America• Georgia Tech (Atlanta)• Johns Hopkins University (Baltimore)• University of California (San Diego)• …
Asia• University of New South Wales (Sydney)• …
GDR BioComp – Colloque 2015 – St Paul de Vence
Thanks for your attention