Genetic Resilience and Variable Mutation Rates

It has been shown that evolutionary computation methods are influenced not only by the fitness function explicitly defined by the user but also by the genetic resilience inherent in the evolutionary mechanisms. Given an environment with specialized ecological niches that allow for high fitness and niches that are less specialized but have a lower maximum fitness, Jones and Soule (2006) investigate the effect on the resilience of individuals of a crossover mechanism that allows chromosomes to increase or decrease in size. They conclude that after the genes gain size resulting in resilience, the al-gorithm performs as if it is influenced only by the fit-ness function. In the present work, we investigate the effect of variant mutations on resilience in evolutionary computation, specifically in genetic algorithms. The core of the experiments study the effects of a mapping function from fitness to mutation rate on the shape of the existing fitness landscape. We determine whether, by manipulating this fitness-to-mutation-rate function, individuals will effectively become more resilient and the population will converge based on fitness alone, given an environment such as the one used by Jones and Soule. 1