Learning classier systems are a machine learning paradigm introduced by John Holland in 1976. They are rule-based systems in which learning is viewed as a process of ongoing adaptation to a partially unknown environment through genetic algorithms and temporal dierence learning.
From the beginning, classier systems have attracted the interest of researchers in many dierent areas, ranging from the design of gas pipelines to personal internet agents, and including cognitive science, data mining, economic trading agents, and autonomous robotics. In 1989 Stewart Wilson and David Goldberg presented a review of the rst decade of classier system research, discussing some of the milestones that characterized the eld's early development. In 1992 the First International Workshop on Learning Classier Systems (IWLCS92) was held in Houston, Texas.
The 1990s saw an increasing interest in the eld: many successful applications to real-world problems were presented as well as new classier system models. With seven years since the rst workshop and more than 400 papers published, it was time in 1999 to examine again the state of the art of learning classier system research. For this purpose the Second International Workshop on Learning Classier Systems (IWLCS99) was organized and the idea of this volume conceived. The workshop, held in Orlando, Florida, July 13, 1999, attracted a vital community of researchers from many dierent areas who share a common interest in this machine learning paradigm. Some of the most interesting work presented at the workshop appears in this volume.
Our book provides an overview of the current state of the art of learning classier systems and highlights some of the most promising research directions. The rst paper of Part I asks the fundamental question: \What Is a Learning Classier System?". Answers are given by John Holland, originator of classier systems, and by other long-time researchers in the eld: Lashon Booker, Marco Colombetti, Marco Dorigo, Stephanie Forrest, David Goldberg, Rick Riolo, Robert E. Smith, and Stewart Wilson. Three following papers summarize, from dierent perspectives, developments in the eld since Wilson and Goldberg's 1989 review. Part II contains papers on advanced topics of current interest, including alternative representations, methods for evaluating rule utility, and extensions to existing classier system models. Part III is dedicated to promising applications of classier systems such as: data mining, medical data analysis, economic trading agents, aircraft maneuvering, and autonomous robotics. A classier systems bibliography with more than 400 references completes the book.
We hope that this volume will be a key reference for researchers working with learning classier systems over the next decade; the constantly increasing interest in this paradigm suggests that there will be many of them!