The dynamical behavior of classifier systems
John H. Miller, Stephanie Forrest · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1989
Classifier systems are quite complicated, in terms of both their components and behavior. This complexity is understandable given the wide spectrum of activity they are intended to model. Unfortunately, the complexity of these systems also makes it difficult to understand them analytically. Previous analysis has focused on specific components of the classifier system, for example, the genetic algorithm or the bucket brigade. The lack of a unified theory has led users of these systems to rely on ad hoc methods for choosing representations and parameter settings. Recent results (Riolo, 1988) indicate that classifier systems can be very sensitive to particular encodings and parameter choices. In this paper, we propose a methodology for studying the interactions among various components of the classifier system architecture.