A self-adaptive mating restriction strategy based on survival length for evolutionary multiobjective optimization

作者：Li, Xin; Zhang, Hu; Song, Shenmin^*

来源：Swarm and Evolutionary Computation, 2018, 43: 31-49.

DOI：10.1016/j.swevo.2018.02.009

• Multiobjective optimization
• Evolutionary algorithm
• Mating restriction
• K-means clustering algorithm
• Differential evolution

Mating restriction plays an important role in improving the performance of the multiobjective evolutionary algorithms. However, research on mating restriction is still insufficient, and in most cases, the same mating restriction strategy is set for all individuals without consideration of individual quality. To effectively alleviate the above problems, this paper proposes a self-adaptive mating restriction strategy based on survival length (MRSL). Based on the information of neighbourhood relationship extracted by clustering, mating parents are selected from the neighbours for exploitation or from the whole population for exploration, which is controlled by the mating restriction probability. Furthermore, considering different individuals are in need of different search space, each individual has a separate mating restriction probability. Additionally, the survival length of the individual is proposed to define and update the mating restriction probability self-adaptively at each generation. The survival length of the individual represents the number of survival generations over the last certain period of time. A multiobjective differential evolutionary algorithm called MDESL is proposed by using MRSL, and compared with five representative algorithms on several test instances. The experimental results validate the superiority of MDESL over the comparison algorithms.

• 出版日期2018-12
• 单位哈尔滨工业大学; 北京市电加工研究所; 北京机电研究所