Multi-objective evolutionary algorithms (MOEAs) can efficiently solve difficult problems when other metaheuristics fail. In order to design a fast MOEA that can find very good solutions, the values of the parameters have to be carefully selected. This talk focuses on the improvement of the anytime behavior of MOEAs by using an automatic configuration tool. The anytime optimization problem is modeled as a bi-objective Pareto front, where the hypervolume development is measured over time. The goal is to find algorithmic configurations that obtain the best Pareto fronts. Three different evolutionary algorithms, called IBEA, NSGAII and SPEA2, have been analyzed by using the iterated racing procedure combined with the hypervolume indicator. The problem instances are successively solved with each MOEA and the results obtained after the parameter tuning are compared with the ones using pre-defined parameters values.
Anytime Behaviour, Multiobjective Evolutionary Algorithms