Emergent gender differentiation in gender genetic algorithm meta-optimization (Group, Ring, Language and Related Areas in Computer Science)

Abstract

Genetic algorithms are metaheuristics first presented in their modern form by Holland in 1992. There are many variations, but all involve an evolving population of candidate solutions to some problem. While most implementations involve asexual reproduction, some attempts have been made to harness advantages of sexual reproduction in genetic algorithms -such approaches are called gender genetic algorithms. In particular, the male mutation bias, which has been well documented in mammals, depends on gender differentiation in parameters relevant to genetic algorithms. We implement gender and perform meta-optimization on the onemax benchmark and statistically analyze whether gender differences in mutation rate and tournament size naturally emerge.