Understanding interactions among genetic algorithm parameters

Abstract

Genetic algorithms (GAs) are multi-dimensional and stochastic search methods, involving complex interactions among their parameters. For last two decades, researchers have been trying to understand the mechanics of GA parameter interactions by using various techniques. The methods include careful 'functional' decomposition of parameter interactions, empirical studies, and Markov chain analysis. Although the complex knot of these interactions are getting loose with such analyses, it still remains an open question in the mind of a new-comer to the field or to a GA-practitioner as to what values of GA parameters (such as population size, choice of GA operators, operator probabilities, and others) to use in an arbitrary problem. In this paper, we investigate the performance of simple tripartite GAs on a number of simple to complex test problems from a practical standpoint. Since function evaluations are most time-consuming in a real-world problem, we compare different GAs for a fixed number of function evaluations. Based on probability calculations and simulation results, it is observed that for solving simple problems (unimodal or small modality problems) mutation operator plays an important role, although crossover operator can also solve these problems. However, two operators (when applied alone) have two different working zones for population size. For complex problems involving massive multimodality and misleadingness (deception), crossover operator is the key search operator and performs reliably with an adequate population size. Based on these studies, it is recommended that when in doubt, the use of the crossover operator with an adequate population size is a reliable approach.