Learning in the Associative Segment
The learning rule used here is the error-correction method. The output of blocks of the second layer of the visual segment are combined to produce the input of the neurons in the associative segment. Let us assume that the input to the associative segment is E = (E1;E2;...;En), where, Ei; i = 1;...; n; is a vector, and the desired output is D = (d1; d2; ...; dl), where l is the number of classes (neurons in the second layer) required to classify them. Assume also that the output of the associative segment is Y = (y1; y2;...; yn), where, yi; i = 1; ...; n; is a vector of length l, and each element, ypj, is given by
Then the learning algorithm can be stated as follows: