Rules and generalization capacity extraction from ANN with GP

Abstract

Different techniques for extracting Artificial Neural Networks (ANN) rules have been used up to the present time, but most of them have focused on certain types of networks and their training. However, there are practically no methods which deal with ANN rule-discovery as systems that are independent from their architecture, training, and internal distribution of weights, connections, and activation functions. This paper proposes a method based on Genetic Programming (GP) with the purpose of achieving the generalization capacity characteristic of ANNs, by means of symbolic rules which can be understood by human beings.