A self-adaptive recombination method in evolutionary algorithms for solving epistatic problems

There are many different forms of recombination operators available in literature. However, it is difficult to determine a priori which one is the best suited for a given problem. This issue encourages us to propose an adaptive evolutionary algorithm to solve the NK landscape problem, which dynamically selects the recombination operator from an operator pool during the evolution; this removes the need of specifying a single recombinator operator ad-hoc. We compare the performance of our adaptive proposal against traditional evolutionary algorithms in a numerical way. Our experiments show that the simple adaptive mechanism has a good performance among all the evaluated ones on high dimensional landscapes with an additional reduction in pretuning time. Presentado en XII Workshop Agentes y Sistemas Inteligentes (WASI) Red de Universidades con Carreras en Informática (RedUNCI)