N1 X
~x~xRðxi Þ ¼kyð j Þ 2 yð j;i Þk2
N j¼1
ð1Þ
Rðxi Þ
N
~xyð j Þ
xji
~xyð ji Þ
relevance of component xi of input vector x;
number of patterns;
output vector of the neural network for each pattern
xi of input presented;
input vector xj ; where ith component was sub-
stituted for its average value based on all the input
vectors;
output of the network for the input xji :