A novel two-archive matching-based algorithm for multi- and many-objective optimization

摘要

In evolutionary multi-objective optimization, it is crucial for the evolutionary algorithm to maintain a good balance between convergence and diversity. The recently proposed Two_Arch2 algorithm provides a new perspective to solve this problem. However, due to the properties of the mating mechanism and the limitations of the distance-based diversity maintenance scheme, both the computational complexity and the diversity face great challenges as the number of objectives increases. In this paper, we propose an improved Two-Archive algorithm for both multi- and many-objective optimization, aiming at further promoting the balance between convergence and diversity. In the proposed algorithm, we introduce a decomposition idea into the mating pool of the convergence archive, which increases the number of favorable solutions and reduces the computational complexity. At the same time, we apply a penalty angle-based selection scheme to the diversity archive, which effectively maintains the population diversity. The effectiveness of the proposed algorithm is compared with five state-of-the-art multi-objective evolutionary algorithms on a variety of benchmark problems. The experimental results demonstrate that the proposed algorithm has highly competitive performance on both multi- and many-objective optimization problems in particular, remedying problems of Two_Arch2. (C) 2019 Published by Elsevier Inc.