Gendered Selection Strategies in Genetic Algorithms for Optimization

The selection operator in the standard genetic algorithm (GA) determines which individuals are chosen from a relatively homologous population for mating and crossover. This operator is crucial for the performance of the GA, since it may lead the algorithm to premature convergence and limited search scope (or genetic diversity) by repeatedly choosing very strong individuals with similar genetic code. In the model proposed here, a sexual strategy is introduced by simulating distinct gender groups, with each gender having different partner selection criteria, and a model of sexual selection that allows for competition between individuals in the same group and co-operation when a mating relation is established. As in natural systems, crossover is only permitted between individuals in contrasting gender groups, and the mutation probabilities depend on the individual’s gender. Experimental results on some standard optimization problems provides evidence that this is a useful strategy.