01.11.2014 Views

MACHINE LEARNING TECHNIQUES - LASA

MACHINE LEARNING TECHNIQUES - LASA

MACHINE LEARNING TECHNIQUES - LASA

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

155<br />

C<br />

µν<br />

−1<br />

where ( )<br />

is the µν th element of the inverse matrix<br />

overlap of each pattern on each bit and is given by:<br />

( C)<br />

µν<br />

N<br />

i=<br />

1<br />

µ ν<br />

i i<br />

1<br />

C − . C measures the degree of<br />

= ∑ x x<br />

(6.72)<br />

The price paid for the improvement on the network capacity is that the learning rule becomes<br />

non-local, which makes the model less attractive from a biological point of view.<br />

6.9.5 Convergence of the Static Hopfield Network<br />

One can show that the Hopfield Network with an asynchronous update rule is bound to converge.<br />

The idea is that the dynamics of the system can be described by an Energy function:<br />

1<br />

E =− ∑ Wijss<br />

(6.73)<br />

i j<br />

2<br />

i,<br />

j<br />

It remains to show that E is a Lyapunov function; i.e., it is bound to converge to a stable fixed<br />

*<br />

point x . A Lyapunov function E is, by definition, a continuous differentiable, real value function<br />

with the following two properties:<br />

* *<br />

( ) ( )<br />

Positive Definite: E x > 0 ∀x≠ x and E x = 0<br />

(6.74)<br />

d *<br />

All trajectory flow downhill: E ( x ) < 0 ∀ x ≠ x<br />

(6.75)<br />

dx<br />

Hint: Once the network has converged, all units have reached a stable state, i.e.<br />

i<br />

( ) ( 1 )<br />

s t = s t+ ∀ i.<br />

i<br />

6.10 Continuous Time Hopfield Network and Leaky-Integrator Neurons<br />

As mentioned previously, the convergence and capacity of the Hopfield network can be majorly<br />

affected when changing from asynchronous update to synchronous update. It turns out that, once<br />

we move to a continuous-time version of the Hopfield network, this issue melts away.<br />

Let us assume that each neuron's activity<br />

x<br />

j<br />

is a continuous function of time xj<br />

( )<br />

t . The neuron's<br />

response to the activation of other neurons in the network is then given by a first-order differential<br />

equation:<br />

© A.G.Billard 2004 – Last Update March 2011

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!