Similarities Between Co-evolution and Learning Classifier Systems and Their Applications

Ramón Alfonso Palacios-Durazo¹ and Manuel Valenzuela-Rendón²

¹Lumina Software, Washington 2825 Pte C.P. 64040, Monterrey N.L., Mexico
apd@luminasoftware.com
http://www.luminasoftware.com/apd

²Centro de Sistemas Inteligentes, Instituto Tecnológico y de Estudios Superiores de Monterrey, Sucursal de Correos J, C.P. 64849, Monterrey, N.L., Mexico
valenzuela@itesm.mx
http://www-csi.mty.itesm.mx/~mvalenzu

Abstract. This article describes the similarities between learning classifier systems (LCSs) and coevolutionary algorithm, and exploits these similarities by taking ideas used by LCSs to design a non-generational coevolutionary algorithm that incrementally estimates fitness of individuals. The algorithm solves some of the problems known to exist in coevolutionary algorithms: it does not loose gradient and is successful in generating an arms race. It is tested on MAX 3-SAT problems, and compared to a generational coevolutionary algorithm and a simple genetic algorithm.

LNCS 3102, p. 561 ff.

Full article in PDF

lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2004