A simple payoff-based learning classifier system

It is now ten years since Wilson introduced the 'Zeroth-level' learning classifier system with the aim of simplifying Holland's original system to both aid understanding and improve performance. Despite being comparatively simple, it is still somewhat complex and more recent work has shown the system's sensitivity to its control parameters, particularly with respect to the underlying fitness sharing process. This paper presents a simple payoff-based learning classifier system with which to explore aspects of fitness sharing in such systems, a further aim being to achieve similar performance to accuracy-based learning classifier systems. The system is described and modelled, before being implemented and tested on the multiplexer task.