对不连续帕累托前沿的一种改进的MOEA/D  

An Improved MOEA/D for Discontinuous Pareto Fronts

作　　者：张宁 高尚[1] ZHANG Ning;GAO Shang(College of Computer Science,Jiangsu University of Science and Technology,Zhenjiang 212003)

机构地区：[1]江苏科技大学计算机学院,镇江212003

出　　处：《计算机与数字工程》2021年第11期2189-2193,共5页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61772244)资助。

摘　　要：提出了一种改进的基于分解的多目标进化算法,用于解决不连续帕累托前沿的多目标优化问题中出现帕累托近似前沿分布不均匀与不完整的问题。主要的思想是通过基于密度的聚类算法将尽量逼近帕累托前沿的种群划分为若干个子种群,将不连续帕累托前沿问题转化为多个连续子问题,然后协同演化所有子种群,最后获得更为均匀与完整的帕累托解集。实验表明对于处理不连续帕累托问题的优越性。An improved decomposition-based multiobjective evolutionary algorithm is proposed to solve the problem of un⁃even and incomplete Pareto approximation in the multiobjective optimization problem of discontinuous Pareto front.The main idea is to divide the population that is as close as possible to the Pareto Front into several sub-populations by density-based clustering algo⁃rithm,transform the discontinuous Pareto frontier problem into multiple contiguous sub-problems.Then coordinately evolve all sub-populations to get a more uniform and complete Pareto solution set.Experiments have shown the superiority of dealing with the discontinuous Pareto problem.

关 键 词：进化算法 多目标优化 密度聚类 分解 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]