PUB - Publikationen an der Universität Bielefeld

A multi-objective multifactorial evolutionary algorithm has been proposed recently to address multi-objective multi-tasks optimization simultaneously. However, the approach only focuses on the improvement of algorithm convergence via knowledge transfer among the optimization tasks. To enhance the performance of both diversity and convergence which are important for evolutionary multi-objective optimization, this paper proposes a two-stage assortative mating method for multi-objective multifactorial evolutionary optimization. In the proposed algorithm, decision variables are first divided into two types using a decision variable clustering method: diversity-related variables and convergence-related variables. The two types of variables then undergo assortative mating with different parameters independently when offspring are generated. Experimental results on a variety of test instances show that the proposed algorithm is highly competitive as compared with existing multi-task and single-task algorithms.