Multiobjective optimization by decomposition with Pareto-adaptive weight vectors
Jiang Siwei, Cai Zhihua, Jie Zhang, Ong Yew-Soon · 2011
MOEA/D is a recently proposed methodology of Multiobjective Evolution Algorithms that decomposes multiobjective problems into a number of scalar subproblems and optimizes them simultaneously. However, classical MOEA/D uses same weight vectors for different shapes of Pareto front. We propose a novel method called Pareto-adaptive weight vectors (paλ) to automatically adjust the weight vectors by the geometrical characteristics of Pareto front. Evaluation on different multiobjective problems confirms that the new algorithm obtains higher hypervolume, better convergence and more evenly distributed solutions than classical MOEA/D and NSGA-II.