Inversion of a neural network via interval arithmetic for rule extraction

Abstract

In this paper we propose a new algorithm for rule extraction from a trained Multilayer Feedforward network. The algorithm is based on an interval arithmetic network inversion for particular target outputs. The types of rules extracted are N-dimensional intervals in the input space. We have performed experiments with four databases and the results are very interesting. One rule extracted by the algorithm can cover 86% of the neural network output and in other cases 64 rules cover 100% of the neural network output. © Springer-Verlag Berlin Heidelberg 2003.