Local Search Based on Genetic Algorithms

Genetic Algorithms have been seen as search procedures that can quickly locate high performance regions of vast and complex search spaces, but they are not well suited for fine-tuning solutions, which are very close to optimal ones. However, genetic algorithms may be specifically designed to provide an effective local search as well. In fact, several genetic algorithm models have recently been presented with this aim. In this chapter, we call these algorithms Local Genetic Algorithms. In this chapter, first, we review different instances of local genetic algorithms presented in the literature. Then, we focus on a recent proposal, the Binary-coded Local Genetic Algorithm. It is a Steady-state Genetic Algorithm that applies a crowding replacement method in order to keep, in the population, groups of chromosomes with high quality in different regions of the search space. In addition, it maintains an external solution (leader chromosome) that is crossed over with individuals of the population.These individuals are selected by using Positive AssortativeMating, which ensures that these individuals are very similar to the leader chromosome. The main objective is to orientate the search in the nearest regions to the leader chromosome. We show an empirical study comparing a Multi-start Local Search based on the binary-coded local genetic algorithmwith other instances of thismetaheuristic based on local search procedures presented in the literature. The results show that, for a wide range of problems, the multi-start local search based on the binary-coded local genetic algorithm consistently outperforms multi-start local search instances based on the other local search approaches.