weight_eval

USE:

[imp] = weight_eval(wij,wjk)

DESCRIPTION:

Interprets the weight matrices of a trained feed-forward neural network. Gives as return the importance (saliency) index for each of the inputs of the neural network. After an idea by Garson G. D.(1991) extended for multi output networks by Tarca A. L.

ARGUMENTS:

The input to hidden and hidden to output matrices, wij(IxJ) and wjk(K,J) respectively. I= # inputs; J= #hiddend nodes; K=# outputs.

VALUES:

imp is the vector of saliency values for the inputs of the network (supposing the network was correctly trained).

EXAMPLES:

//suppose wij and wjk bellow are the neural network weights. The network configuration is therefore I=3,J=4;K=2

> wij=[4.82 -27.6 13.7 ;15.1  -4.21 -7.58 ; -6.38 -30.5 -12 ; -24.9 -11.2 -11.4 ]';

> wjk=[-1.32 -2.69 -2.56 -2.29; 5.1 –2.2 3.1 –0.3];

[importance] = weight_eval(wij,wjk)

COMMENTS:

Created by Laurentiu Adi Tarca. Revised 10.07.2003

This help was created on 12.03.2004