You are on page 1of 20

Hopfield Neural Network

Presented by:V.Bharanigha
10MEPE06

FEEDBACK/RECURRENT
NETWORK
One

output vector is assigned to every input


vector
Returns back the output to the inputiterative process
Types:
1)Simulated annealing
2)Boltzmann machine
3)Hopfield net

TYPES OF HOPFIELD NET


Discrete

hopfield net

Continuous

hopfield net

DISCRETE HOPFIELD NET


John

Hopfield (1982)

Very simple network

Solution for optimization problems

Have n neurons networked with each other

HOPFIELD NETWORK
An

interconnected network
Every node is connected to every other node

To

If the weight is 0, the connection doesnt matter

use the network, set the values of the


nodes and let the nodes adjust their values
according to the weights.
All the diagonal elements of weight matrix are
zero
Wij=Wji
Wii=0

NEURONS IN HOPFIELD NETWORK


The

neurons are binary units

They are either active (1) or passive(0)


Alternatively + or

The

network contains N neurons


The state of the network is described as a
vector of 0s and 1s

ARCHITECTURE
The

network is fully interconnected

x input and y output neurons


The connections are bidirectional and symmetric
Wi , j W j ,i
The setting of weights depends on the
application

ARCHITECHTURE OF HOPFIELD NET(DISCRETE)

TRAINING ALGORITHM
Applied

for both binary and bipolar vector


patterns
Weight of the matrix is determined by
HEBBs rule
Stores the set of binary input patterns s(p) for
p=1,P
where s(p)={s1(p),si(p),sn(p)}

Contd..,
The weight matrix is given by,
For binary input patterns,
for ij and wii=0

For

bipolar input patterns,


for ij and wii=0

APPLICATION ALGORITHM
Step

1: Initialize weights to store pattern


using hebbs rule
Step 2: For each input vector x, repeat steps
3 to 7
Step 3; Set initial activations of the net equal
to the external input vector x,yi=xi (i=1,n)
Step 4:Perform steps 5 to 7 for each unit y i

Step

5:Compute the net input,

yini
Step

6:Determine output signal,


1,

y j yi ,
0,

if yini i ;
if yini i ;
if yini i ;

Step

7:Broadcast the value of yi to all other

units
Step 8:test for convergence
The value of threshold i is taken to be zero
Each unit is randomly updated at same
average rate

ANALYSIS
Storage

capacity

Accuracy,

for binary patterns

for bipolar patterns

ENERGY CAPACITY
Energy

function

Change

neuron,

in energy due to change in energy of

If

yi is positive , E<0 and yi will be zero if

If

yi is zero, E<0 and yi will be positive if

For

both the cases,the energy cannot


increase i.e. E<0
Hence, the energy is bounded and reaches a
stable equilibrium
Energy change depends only on activation of
one unit

Example:

THANK YOU

You might also like