In this paper we establish a mathematical framework in which we develo
p measures for determining the contribution of individual features to
the performance of a classifier. Corresponding to these measures, we d
esign metrics that allow estimation of the importance of features for
a specific multi-layer perceptron neural network. It is shown that all
measures constitute lower bounds for the correctness that can be obta
ined when the feature under study is excluded and the classifier rebui
lt. We also present a method for pruning input nodes from the network
such that most of the knowledge encoded in its weights is retained. Th
e proposed metrics and the pruning method are validated with a number
of experiments with artificial classification tasks. The experiments i
ndicate that the metric called replaceability results in the tightest
error bounds. Both this metric and the metric called expected influenc
e result in good rankings of the features. (C) 1998 Elsevier Science L
td. All rights reserved.