Enhancing recombination with the Complementary Surrogate Genetic Algorithm

In traditional genetic algorithm (GA) approaches using finite populations, recombination alone has been shown to be insufficient to guarantee optimal solutions because of the well known problems of fixation of alleles and premature convergence. Mutation is widely regarded as critical to preserve diversity in recombination dominant GAs, as well as a powerful search heuristic in its own right; mutation is central to recent GA convergence proofs. The paper examines an alternate genetic algorithm with no explicit mutation operator. The Complementary Surrogate GA (CSGA) uses traditional crossover operators, but guarantees recombination access to the complete search space by modifying the GA population structure. Complementary Surrogate Sets (CSS) within the population ensure allele diversity at each locus, while allowing standard selection methods to work as expected. A proof of convergence is provided as well as the results of an empirical study examining the CSGA using various CSS strategies on standard function optimization benchmarks