Getting Lost or Getting Trapped: On the Effect of Moves to Incomparable Points in Multiobjective Hillclimbing

摘要

Divergent behavior may occur in elitist multiobjective EAs which allow moves to incomparable solutions. We study under which conditions this is exhibited. For simple model landscapes stochastic dynamics are studied and quantified by means of Markov chains. The studies suggest that increasing the population size tempers divergent behavior. In addition, we study whether common elitist algorithms such as NSGA-II and SMS-EMOA have divergent behavior.