Illinois Genetic Algorithms Laboratory

Moreno-Torres, J. G., Llorà, X., Goldberg, D. E. (2009)
TR No.: 2009003 | Download PDF | Download PS

Abstract: One of the main problems that arises when using gene expression programming conditions in learning classifier systems is the increasing number of symbols present as the problem size grows. This issue severely limits the scalability of the technique, due to model building becoming untractable. We propose a binary representation of GEP chromosomes as a possible solution. We provide a theoretical reasoning behind our representation, and back it up with empirical results.