基于极大极小距离密度的多目标微分进化算法  

A Multi-Objective Differential Evolution Algorithm Based on Max-Min Distance Density

作　　者：张利彪[1] 周春光[1] 马铭[1] 孙彩堂[1] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012

出　　处：《计算机研究与发展》2007年第1期177-184,共8页Journal of Computer Research and Development

基　　金：国家自然科学基金重点项目(60433020);吉林大学"九八五"工程"计算与软件科学"科技创新平台基金项目;教育部"符号计算与知识工程"重点实验室基金项目(02090)

摘　　要：微分进化(differential evolution)是一种新的简单而有效的直接全局优化算法,并在许多领域得到了成功应用.提出了基于极大极小距离密度的多目标微分进化算法.新算法定义了极大极小距离密度,给出了基于极大极小距离密度的Pareto候选解集的维护方法,保证了非劣解集的多样性.并根据个体间的Pareto支配关系和极大极小距离密度改进了微分进化的选择操作,保证了算法的收敛性,实现了利用微分进化算法求解多目标优化问题.通过对5个ZDT测试函数、两个高维测试函数的实验及与其他多目标进化算法的对比和分析,验证了新算法的可行性和有效性.Differential evolution is a simple and powerful globally optimization new algorithm. It is a population-based, direct search algorithm, and has been successfully applied in various fields. A multi- objective differential evolution algorithm based on max-min distance density is proposed. The algorithm proposed defines max-min distance density and gives a Pareto candidate solution set maintenance method, ensuring the diversity of the Pareto solution set. Using Pareto dominance relationship among individuals and max-min distance density, the algorithm improves the selection operation of differential evolution, ensures the convergence of the algorithm, and realizes the solution of multi-objective optimization problems by differential evolution. The proposed algorithm is applied to five ZDT test functions and two high dimension test functions, and it is also compared with other multi-objective evolutionary algorithms. Experimental result and analysis show that the algorithm is feasible and efficient.

关 键 词：微分进化 极大极小距离密度 多目标优化问题 多目标进化算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]