Building anticipations in an accuracy-based learning classifier system by use of an artificial neural network

Learning classifier systems which build anticipations of the expected states following their actions are a focus of current research. This paper presents a mechanism by which to create learning classifier systems of this type, here using accuracy-based fitness. In particular, we highlight the supervised learning nature of the anticipatory task and amend each rule of the system with a traditional artificial neural network. The system is described and shown able to perform well in a number of well-known maze tasks.