一种基于邻域的多目标进化算法  

Multi-objective evolutionary algorithm based on neighborhood

作　　者：李密青[1] 郑金华[1] 罗彪[1] 伍军[1] 文诗华[1] 

机构地区：[1]湘潭大学信息工程学院,湖南湘潭411105

出　　处：《计算机应用》2008年第6期1570-1574,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60773047);国家863计划项目(2001AA114060);留学回国人员科研启动基金资助项目(教外司留[2005]546号);湖南省自科基金资助项目(05JJ30125);湖南省教育厅重点科研资助项目(06A074)

摘　　要：种群维护是多目标进化算法的重要组成部分。针对维护方法和运行效率的矛盾,提出一种基于邻域的多目标进化算法(NMOEA)。定义了一个反映个体之间邻近程度的指标———邻域包含关系,利用此关系对个体进行分布适应度分级的赋值,并用动态方法快速地对种群进行维护。通过7个测试问题和3个方面的测试标准,结果表明新算法在较快速地接近真实的最优面的同时,拥有良好的分布性。Population maintenance is an important issue in multi-objective evolutionary algorithms. For the deficiency that the maintenance methods of good distribution usually have a high time complexity, a multi-objective evolutionary algorithm based on neighborhood （named NMOEA） was proposed. This measure defined a criterion-neighborhood containing relation, which represented the close degree of individuals. And it was used to assign diversity fitness in a dynamic method that maintained the population rapidly. By examining three performance metrics on seven test problems, the new algorithm can approach the true Pareto front fast, and has good distribution.

关 键 词：多目标进化算法 多目标优化问题 种群维护 分布适应度 邻域 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]