 |
    |
Shake Them All!
Rethinking Selection and Replacement in MOEA/D
Gauvain Marquet1,2, Bilel Derbel1,2, Arnaud Liefooghe1,2, and El-Ghazali Talbi1,2
1Université Lille 1, LIFL, UMR CNRS 8022, France
2Inria Lille - Nord Europe, DOLPHIN project-team, France
firstname.lastname@inria.fr
Abstract. In this paper, we build upon the previous efforts to enhance the search ability of Moea/d (a multi-objective decomposition-based algorithm), by investigating the idea of evolving the whole population simultaneously. We thereby propose new alternative selection and replacement strategies that can be combined in different ways within a generic and problem-independent framework. To assess the performance of our strategies, we conduct a comprehensive experimental study on bi-objective combinatorial optimization problems. More precisely, we consider 
MNK-landscapes and knapsack problems as a benchmark, and experiment a wide range of parameter configurations for Moea/d and its variants. Our analysis reveals the effectiveness of our strategies and their robustness to parameter settings. In particular, substantial improvements are obtained compared to the conventional Moea/d.
LNCS 8672, p. 641 ff.
lncs@springer.com
© Springer International Publishing Switzerland 2014