from brain activities to mathematical models

25
From brain activities to mathematical models The TempUnit model, a study case for GPU computing in scientific computation.

Upload: howell

Post on 05-Jan-2016

28 views

Category:

Documents


1 download

DESCRIPTION

From brain activities to mathematical models. The TempUnit model, a study case for GPU computing in scientific computation. What part of the brain?. How to study it ?. First attempt: use of a MLP. What is a MLP?. First Attempt: MLP (2). Results (1). Results (2). Crack the code !!. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: From brain activities to mathematical models

From brain activities to mathematical models

The TempUnit model, a study case for GPU computing in

scientific computation.

Page 2: From brain activities to mathematical models
Page 3: From brain activities to mathematical models

MN

CM

IN

Muscle

Cortex

Moelle

Périph:MuscleMouvement

What part of the brain?

Page 4: From brain activities to mathematical models

How to study it ?

Page 5: From brain activities to mathematical models

First attempt: use of a MLP• What is a MLP?

Page 6: From brain activities to mathematical models

First Attempt: MLP (2)

Page 7: From brain activities to mathematical models

Results (1)

Page 8: From brain activities to mathematical models

Results (2)

Page 9: From brain activities to mathematical models

Crack the code !!• Frequency code (Number of spikes in a time

lap) ?• Spatial coding (distributed trough the

network) ?• Temporal code (Precise binary pattern) ?• Spatio-temporal code (Synchronies) ?• Something else ?

Page 10: From brain activities to mathematical models
Page 11: From brain activities to mathematical models

The modelx

Xt

Page 12: From brain activities to mathematical models

Learn the parameters vi

• Solving a system of linear equation oversized.• Much faster and straightforward than

backpropagation for the MLPExample of a learned basis function

Page 13: From brain activities to mathematical models

Performances compared to MLP

Page 14: From brain activities to mathematical models

Check Chap. 12

Page 15: From brain activities to mathematical models

Graph of Neuronal Activity• The output activity of a TempUnit neural

network can be described by a graph directly related to its connectivity– You determine the topology of your graph easily

• Allow to determine the input activity for a particular desired output

Page 16: From brain activities to mathematical models

Can a real biological neuron do that ?

Page 17: From brain activities to mathematical models
Page 18: From brain activities to mathematical models
Page 19: From brain activities to mathematical models

Pattern recognition

Page 20: From brain activities to mathematical models
Page 21: From brain activities to mathematical models

learning rules for unsupervised learning

Page 22: From brain activities to mathematical models

EPSP from the integrate-and-fire model

0

r

ax

s

ax t

rsr

t

rs

ek

ek

To find the position of the maximum (peak), one has to resolve the following equation:

ax

tt

rs

teek

tu r

ax

s

ax

)(

(Gerstner & Kistler, 2002)

rsax

r

sax

r

s

t

ln

1

1

From the integrate and fire, the α function:

time

Page 23: From brain activities to mathematical models

The new equation of the TempUnit model:

• With μ, the maximum value:

r

axrs

ax

r

sax

r

s

sax

rsax

r

sax

r

s

eers

ln

1

1ln

1

1

1

v

pt

eep

kptu

saxv

pt

v

pt

srs

r

sax

s

sax

1),(

time

p=0

p=6

p : position of the synapse

Page 24: From brain activities to mathematical models

From equations to a simulation software

Page 25: From brain activities to mathematical models