基于IGD+指标的两阶段选择高维多目标进化算法

IGD+ indicator based many-objective evolutionary algorithm

中文关键词  高维多目标优化  IGD+指标  两阶段选择策略  参考点分布自适应策略  种群分解策略  进化算法

英文关键词  many-objective optimization  IGD+indicator  two-stage selection strategy  adaptive strategy of reference point distribution  population partition strategy  evolutionary algorithm

中文摘要

      针对在高维空间下多目标进化算法难以维持种群收敛性和多样性平衡的问题, 本文提出一个基于IGD+指标的两阶段选择高维多目标进化算法(MaOEA–ITS). 在第1阶段, 算法基于IGD+指标选择收敛性良好的精英个体, 其所需的参考点通过引入切割平面截距法构建. 在第2阶段, MaOEA–ITS使用模糊c均值算法对参考向量进行聚类, 聚类后的参考向量引导种群分解策略对剩余个体进行环境选择, 从而维持种群的多样性. 另外, 为了保护能够提高种群多样性的极值解, 本文提出一个参考点分布自适应策略. 最后, 通过仿真实验来验证MaOEA–ITS的有效性和优越性.

英文摘要

      In order to solve the problem that the multi-objective evolutionary algorithm is difficult to balance between the population convergence and diversity in high-dimensional space, in this paper, an IGD+indicator based many-objective evolutionary algorithm with two stage selection (MaOEA–ITS) is proposed. In the primary stage, the proposed algorithm adopts IGD+indicator as selection criterion to select individuals with favourable convergence, and the reference points required are constructed by introducing the intercepts way of cutting plane. In the second stage, the MaOEA–ITS uses a fuzzy c-means algorithm to cluster reference vectors. Then, reference vectors clustered guide population partition strategy to select remaining individuals of population, thereby maintaining the population diversity. In addition, for protecting the extreme solutions that enables to improve population diversity, a reference point distribution based on the adaptive strategy is proposed. Finally, simulation experiments are used to verify the effectiveness and superiority of MaOEA–ITS.