基于聚类的昂贵多目标优化代理辅助进化算法  

A clustering-based surrogate-assisted evolutionary algorithm for expensive multi-objective optimization

作　　者：白富生[1] 陈姣伶 BAI Fusheng;CHEN Jiaoling(National Center for Applied Mathematics in Chongqing,Chongqing Normal University,Chongqing 401331,China;School of Mathematical Sciences,Chongqing Normal University,Chongqing 401331,China)

机构地区：[1]重庆师范大学重庆国家应用数学中心,重庆401331 [2]重庆师范大学数学科学学院,重庆401331

出　　处：《运筹学学报》2022年第4期31-42,共12页Operations Research Transactions

基　　金：国家自然科学基金(Nos.11991024,11871128);重庆自然科学基金(No.cstc2019jcyjmsxmX0368);重庆市技术创新与应用发展专项重点项目(No.cstc2021jscx-jbgsX0001);重庆市教委科学技术研究计划重点项目(No.KJZD-K202114801)。

摘　　要：针对目标函数估值昂贵的多目标优化问题,提出了基于聚类的代理辅助进化算法。在MOEA/D算法的框架下,对种群进行聚类,并通过权重向量的邻域选出种群子集,在子集上使用径向基插值函数辅助的差分进化算法得到新解,对种群进行更新。在7个DTLZ标准测试问题上进行了数值实验,计算结果表明本文提出的算法比新近提出的多目标邻域回归优化(MONRO)算法具有优势。A clustering-based surrogate-assisted evolutionary algorithm is proposed for computationally expensive multi-objective optimization problems.Under the framework of MOEA/D,the population is partitioned into several clusters,and the population subsets are formed via the neighbourhood of the weights.Then the radial basis function surrogate-assisted differential evolution algorithm is used to generate new solution points from the formed subsets,and the population is updated using the generated new solution.Numerical experiments have been undertaken on 7 DTLZ test problems,and the computational results indicate that the proposed evolutionary algorithm has advantages over the newly developed multi-objective neighborhood regression optimization(MONRO)algorithm.

关 键 词：多目标优化 代理辅助进化算法 径向基函数 聚类 

分 类 号：O221.6[理学—运筹学与控制论] O221.2[理学—数学]