Rule Extraction from Neural Networks Formed Using a Random Optimization Method with Deterministic Mutation

Abstract

This paper presents a method of extracting rules from neural networks trained using a random optimization method (ROM) with deterministic mutation (DM). The DM is performed on the basis of the result of neural network structure learning. The ROM with the DM is utilized to reduce the number of network connections for iris data. The network connections survived after training represent rules to perform pattern classification for the iris data. The rules are then extracted from the neural network in which hidden units use signum output functions to produce binary values. It enables us to extract simple logical functions from the network. Simulation results for the iris data show this method can generate simple rules compared with conventional methods.