Free Access
Issue
E.J.E.S.S.
Volume 14, Number 1, 2000
Neural Models in Economy and Management Science
Page(s) 1 - 16
DOI https://doi.org/10.1051/ejess:2000104
DOI: 10.1051/ejess:2000104

European Journal of Economic and Social Systems 14 N$^\circ$ 1 (2000) 1-16

Pruning neural networks by minimization of the estimated variance

Peter Morgan, Bruce Curry and Malcom Beynon

Cardiff Business School, Aberconway Building, Colum Drive, Cardiff, CF1 3EU, Wales, UK

Abstract:

This paper presents a series of results on a method of pruning neural networks. An approximation to the estimated variance of errors, V, is constructed containing a supplementary parameter, a - the estimated variance itself being the limit of the function, V, as a tends to zero. The network weights are fitted using a minimization algorithm with V as objective function. The parameter, a, is reduced successively in the course of fitting. Results are presented using synthetic functions and the well-known airline passenger data. We find, for example, that the network can discover, in the course of being pruned, evidence of redundancy in the variables.

Keywords: Neural network, pruning, generalization, penalty function,
estimated variance.

Copyright EDP Sciences 2000