An Accuracy-based Neural Classifier System

Learning Classifier Systems have traditionally used a binary representation, with wildcards added to facilitate generalization. As they are applied to more complex domains the simple representation can become limiting. In this paper we present results from the use of a neural network-based representation scheme within the accuracy-based XCS. Here each rule’s condition and action are represented by a small neural network, evolved through the actions of the genetic algorithm. After describing the changes required to the standard production system functionality, optimal performance is presented for both single-step and multi-step tasks before its extension for function approximation is described.