[en] In this article we introduce a new multiobjective optimizer based on a recently proposed metaheuristic algorithm named Variable Mesh Optimization (VMO). Our proposal (multiobjective VMO, MOVMO) combines typical concepts from the multiobjective optimization arena such as Pareto dominance, density estimation and external archive storage. MOVMO also features a crossover operator between local and global optima as well as dynamic population replacement. We evaluated MOVMO using a suite of four wellknown benchmark function families, and against seven state-of-the-art optimizers: NSGA-II, SPEA2,MOCell, AbYSS,SMPSO,MOEA/DandMOEA/D.DRA. The statistically validated results across the additive epsilon, spread and hypervolume quality indicators confirm that MOVMO is indeed a competitive and effective method for multiobjective optimization of numerical spaces.
Disciplines :
Computer science
Author, co-author :
Salgueiro, Yamisleydi; University of Las Tunas > Faculty of Technical Sciences, Dept. of Informatics Sciences
Toro Pozo, Jorge Luis ; University of Luxembourg > Faculty of Science, Technology and Communication (FSTC) > Computer Science and Communications Research Unit (CSC)
Bello, Rafael; Central University of Las Villas > Computer Science Department
Falcon, Rafael; University of Ottawa > Electrical Engineering and Computer Science