A study on non-random mating and varying population size in genetic algorithms using a royal road function

Abstract

We present a study on the effects of non-random mating and varying population size in genetic algorithm (GA) performance. We tested two algorithms: the non-incest genetic algorithm with varying population size (niGAVaPS) and the negative assortative mating genetic algorithm with varying population size (nAMGAVaPS), on a royal road function. These algorithms mimic natural species behavior by selecting parents according to parenthood (niGAVaPS) or phenotype similarity (nAMGAVaPS). We show that both algorithms outperform simple GA in the example shown. The results suggest that this may be due to the fact that genetic diversity is kept at a higher level by niGAVaPS and nAMGAVaPS, preventing the premature convergence of the algorithms to local optima.