Neural Computation, Vol 9, 185-204, Copyright © 1997 by The MIT Press

LETTERS

A Penalty-Function Approach for Pruning Feedforward Neural Networks

Rudy Setiono

This article proposes the use of a penalty function for pruning feedforward neural network by weight elimination. The penalty function proposed consists of two terms. The first term is to discourage the use of unnecessary connections, and the second term is to prevent the weights of the connections from taking excessively large values. Simple criteria for eliminating weights from the network are also given. The effectiveness of this penalty function is tested on three well-known problems: the contiguity problem, the partiy problems, and the monks problems. The resulting pruned networks obtained for many of these problems have fewer connections than previously reported in the literature.

This article has been cited by other articles:

				B. Baesens, R. Setiono, C. Mues, and J. Vanthienen Using Neural Network Rule Extraction and Decision Tables for Credit-Risk Evaluation Management Science, March 1, 2003; 49(3): 312 - 329. [Abstract] [PDF]

				Z. Chen and S. Haykin On Different Facets of Regularization Theory Neural Comput., December 1, 2002; 14(12): 2791 - 2846. [Abstract] [Full Text]

				D. E. Duckro, D. W. Quinn, and S. J. Gardner III Neural Network Pruning with Tukey-Kramer Multiple Comparison Procedure Neural Comput., May 1, 2002; 14(5): 1149 - 1168. [Abstract] [Full Text]

				R. Setiono Feedforward Neural Network Construction Using Cross Validation Neural Comput., December 1, 2001; 13(12): 2865 - 2877. [Abstract] [Full Text]