An open issue in multi-objective optimization is designing metaheuristics that reach the Pareto front using a low number of function evaluations. In this paper, we adopt a benchmark composed of three well-known problem families (ZDT, DTLZ, and WFG) and analyze the behavior of six state-of-the-art multi-objective metaheuristics, namely, NSGA-II, SPEA2, PAES, OMOPSO, AbYSS, and MOCell, according to their convergence speed, i.e., the number of evaluations required to obtain an accurate Pareto front. By using the hypervolume as a quality indicator, we measure the algorithms converging faster, as well as their hit rate over 100 independent runs. Our study reveals that modern multi-objective metaheuristics such as MOCell, OMOPSO, and AbYSS provide the best overall performance, while NSGA-II and MOCell achieve the best hit rates. © 2008 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Nebro, A. J., Durillo, J. J., Coello Coello, C. A., Luna, F., & Alba, E. (2008). A study of convergence speed in multi-objective metaheuristics. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5199 LNCS, pp. 763–772). https://doi.org/10.1007/978-3-540-87700-4_76
Mendeley helps you to discover research relevant for your work.