Rule-Based Evolutionary Online Learning Systems

Rule-based evolutionary online learning systems, known as Michigan-style learning classifier systems (LCSs), combine reinforcement learning with genetic algorithms for flexible online learning. Despite initial successes, challenges in understanding their complexity and performance across various problem types persisted. This book aims to develop a facetwise theoretical approach to enhance the XCS classifier system, establish a fundamental learning theory, and identify key advantages and applications of LCSs. The analysis reveals competitive machine learning capabilities and facilitates the design of advanced systems.