nueral

Algorithm:
Step0: initialize wts using either the Hebb rule or the Delta rule.
Step1: For each input vector do step 2-4.
Step2: Set activations for input layer units equal to the current vector
Xi = Si
Step3: Compute the net input to the output units
Y-mj = ? Xi Wi j
Step4:Determine the activation of the output units for bipolar targest
1 if y – mj > 0
Yj = 0 if y- mj = 0
-1 if y – mj < 0
The output vector y gives the pattern associated with the input vector x .The hetero associative memory is not iterative.
If the responses of the net are binary , a suitable activation fn is given by
1 if X > 0
F(x)=
0 if X <= 0

An example :-
(Hetroassociatwe net using the Hebb rule )suppose the net is to be trained to store the following mapping from input bipolar vectors S= (S1, S2, S3, S4)
to output vectors with two components t = (t1,t2)
where

S(1) = (1, -1, -1, -1) t (1) = (1 , -1)
S(2) = (1, 1, -1, -1) t(2) = (1, -1)
S(3) = (-1, -1, -1, 1) t (3) = (-1, 1)
S(4) = (-1, -1, 1, 1) t (4) = (-1, 1)

The weight matrix is
W? = ?? Si (?) tj ( ?)
Hence

4 -4
W = 2 -2
-2 2
-4 4


Where the weight matrix to store the first potties pais is given by the outer product of the vectors s = (1, -1, -1, -1) and t = (1, -1) , the weight matrix is 1 -1 1
-1 1 -1 = -1 1
-1 -1 1
-1 -1 1