Self-Adaptation Techniques Applied to Multi-Objective Evolutionary Algorithms


In spite of the success of evolutionary algorithms for dealing with multi-objective optimization problems (the so-called multi-objective evolutionary algorithms (MOEAs)), their main drawback is the fine-tuning of their parameters, which is normally done in an empirical way (using a trial-and-error process for each problem at hand), and usually has a significant impact on their performance. In this paper, we present a self-adaptation methodology that can be incorporated into any MOEA, in order to allow an automatic fine-tuning of parameters, without any human intervention. In order to validate the proposed mechanism, we incorporate it into the NSGA-II, which is a well-known elitist MOEA and we analyze the performance of the resulting approach. The results reported here indicate that the proposed approach is a viable alternative to self-adapt the parameters of a MOEA.