A Collaborative Resource Allocation Strategy for Decomposition-Based Multiobjective Evolutionary Algorithms

Autor: Li Li, Qi Kang, MengChu Zhou, Xinyao Song
Rok vydání: 2019
Předmět:
Zdroj: IEEE Transactions on Systems, Man, and Cybernetics: Systems. 49:2416-2423
ISSN: 2168-2232
2168-2216
DOI: 10.1109/tsmc.2018.2818175
Popis: Decomposition of a multiobjective optimization problem (MOP) into several simple multiobjective subproblems, named multiobjective evolutionary algorithm based on decomposition (MOEA/D)-M2M, is a new version of multiobjective optimization-based decomposition. However, it fails to consider different contributions from each subproblem but treats them equally instead. This paper proposes a collaborative resource allocation (CRA) strategy for MOEA/D-M2M, named MOEA/D-CRA. It allocates computational resources dynamically to subproblems based on their contributions. In addition, an external archive is utilized to obtain the collaborative information about contributions during a search process. Experimental results indicate that MOEA/D-CRA outperforms its peers on 61% of the test cases in terms of three metrics, thereby validating the effectiveness of the proposed CRA strategy in solving MOPs.
Databáze: OpenAIRE