Illinois Genetic Algorithms Laboratory

Tharakunnel, K. K., Butz, M. V., Goldberg, D. E. (2002)
TR No.: 2002022 | Download PDF | Download PS

Abstract:
There has been much work in the past few years exhibiting success of the accuracy based XCS classifier system. At the same time, though, there have been other results that suggest the ill-performance of XCS. Although several studies on XCS are available there is little work explaining the negative results. This paper investigates one particular set of hard problems for XCS. A fundamental limitation of accuracy based fitness in combination with selectorecombinative genetic algorithms (GAs) is revealed.