Improve and Compact Population in XCSFCA using Polynomial Equation
Saeid Goodarzian, Ali Hamzeh, Sattar Hashemi · 2012
XCS is a rule-based evolutionary online learning system. XCSFCA is an extension of XCS where compute continuous actions directly from input states. In XCSFCA, computed actions of a classifier, demonstrated as a straight lines. But in very problems, the desired best action curves are not linear and there are arched; therefore a system with linear action computation needs a large population. This paper studies a new method for compute continuous actions directly from input states. In new proposed method action computes by polynomial equation. Consequently, each classifier represents a nonlinear action curve and the classifiers are more generalized. In comparison with XCSFCA, our method proves to be more efficient and smaller population size.