A Matrix Adaptation Evolution Strategy Based Evolution Algorithm for Large-scale Many-objective Optimization

In recent years, multi-objective optimization problems have been received a lot of attention. However, many applications involve hundreds or thousands of decision variables, which pose a serious barrier for evolutionary algorithms. In this paper, therefore, a novel algorithm, denoted as matrix adaptation evolution strategy based evolution algorithm (MAES-EA), is proposed for large-scale multi-objective optimization. In the proposed algorithm, the decision variables are divided into convergence related variables and diversity related variables, which are optimized by different optimization strategies. To be specific, the matrix adaptation evolution strategy is introduced to reduce the amount of evaluations when optimizing the convergence related variables. In addition, the L
_0.5
-norm-based distance is applied to optimize diversity related variables. The performance of the proposed algorithm is compared with some state-of-the-art methods on LSMOP test instances, and results show that MAES-EA is suitable for large-scale multi-objective optimization.