文献类型：期刊文献

英文题名：A hybrid evolutionary algorithm based on integrated molecular-gradient search for large-scale many-objective optimization

作者：Ding, Weichao[1];Zhang, Tianyu[1];Gu, Chunhua[1];Luo, Fei[1];Dong, Wenbo[1]

机构：[1]East China Univ Sci & Technol, Sch Informat Sci & Engn, 130 Meilong Rd, Shanghai 200237, Peoples R China

年份：2026

卷号：199

外文期刊名：APPLIED SOFT COMPUTING

收录：;EI(收录号：20261920660094);WOS:【SCI-EXPANDED(收录号:WOS:001766033600001)】；

基金：This work is sponsored by the National Natural Science Foundation of China (No. 62403201 and No. 62506130) , Natural Science Foundation of Shanghai (24ZR1415200, 23ZR1414900, 22ZR1416500) .

语种：英文

外文关键词：Conjugate gradient; Molecular collision operator; Evolutionary computation; Large-scale many-objective optimization

摘要：Large-scale many-objective optimization problems (LSMOPs) are widely used in practical applications, which need to simultaneously optimize multiple conflicting objectives under hundreds or thousands of decision vari ables. Existing many-objective evolutionary algorithms (MOEAs) typically address the dimensional curse of LSMOPs by employing techniques such as collaborative evolution, decision variable analysis, search space re duction and operator improvement to decompose large-scale problems, so as to improve algorithm efficiency and performance. However, these approaches often suffer from low computational efficiency due to excessive function evaluations and struggle with intensifying conflicts between convergence and diversity as the number of objectives grows. To tackle the aforementioned challenges, this paper introduces a hybrid large-scale multi-objective evolutionary algorithm that integrates conjugate gradient method with the molecular collisions, with the objective of efficiently solving LSMOPs while preserving the high quality of the solution set. The algorithm adopts a dual-population coevolutionary framework and realizes independent and directed optimization of con vergence and diversity through the organic combination of conjugate gradients and molecular collisions. In order to realize complementary advantages between populations, this algorithm employs a restricted mating selection strategy to generate mating parents and utilizes an independent archive to retain promising solutions during the iteration process. Meanwhile, convergence and diversity enhancement strategies are adopted to maintain the archive. Experimental results on a variety of benchmark test problems show that the proposed algorithm is highly competitive in comparison with existing state-of-the-art MOEAs for LSMOPs.