求解大规模全局优化问题的新型三层递归差分分组方法

New three-level recursive differential grouping method for large-scale optimization problems

中文关键词  全局优化  协同进化  分解方法  三层递归差分分组  递归搜索

英文关键词  global optimization  cooperative coevolution  decomposition methods  three-level recursive differential grouping  recursive search

中文摘要

      协同进化算法在求解大规模全局优化问题上具有较好的效果, 其核心思想是利用分而治之的策略将一个高维问题分解成若干个子问题, 然后分别优化每个子问题. 然而, 现有的分解方法通常需要花费大量的计算成本来获得精确的变量分组. 通过采用递归交互检测中的历史信息简化分组过程, 能够避免检测某些集合的相互关系, 本文提出了一种新型三层递归差分分组策略(NTRDG). 与其他4种现有的分组方法相比, NTRDG在不影响分组精度的情况下计算成本消耗较低. 仿真结果表明, NTRDG在求解大规模全局优化问题时具有很强的竞争力.

英文摘要

      The cooperative coevolution algorithm performs well in solving large-scale global optimization problems. The core idea of cooperative coevolution is to utilize a divide-and-conquer strategy for decomposing high-dimensional problems into multiple subproblems, which are then processed individually and separately. However, existing decomposition methods typically require significant computational cost to obtain accurate variable grouping. To address this issue, a novel three-level recursive differential grouping strategy (NTRDG) is proposed in this paper, which simplifies the grouping process by utilizing historical information in recursive interaction detection and avoids the detection of relationships among certain sets, leading to a lower computational cost without sacrificing grouping accuracy. Simulation results demonstrate that compared to four existing methods, NTRDG exhibits a stronger competitiveness in solving large-scale global optimization problems.