摘要
进化算法具有本质上并行、不需要求导或其他辅助知识、一次运行产生多个解和简单易于实现等优点,被视为求解多目标优化问题的有效方法,目前已经形成了各种不同的多目标进化算法(MOEA)。本文首先回顾了多目标进化算法的研究起源,给出了多目标优化问题的数学描述;其次,详细分析了第一代多目标进化算法,其主要特征是简单易于实现,包括NSGA、NPGA、MOGA等,并指出这一代算法研究的成绩与不足;然后,对第二代多目标进化算法作了全面分析,指出其特征是强调效率,以精英保留策略为实现机制,且对SPEA、PAES、NSGAII、NPGA2、PESA、Micro-GA等方法进行分析比较,还对这一代的研究作了总结;最后,对多目标进化算法的研究趋势作了展望和预测。
Evolutionary Algorithms （EAs） have become popular in multi-objective optimization problems, which are parallel in nature and don＇t require differentiability of objective functions and constraints, and also which deal with a set of possible solutions in a single furl Many Multi-Objective Evolutionary Algorithms （MOEAs） are proposed at present. Firstly, this paper reviews the origin of MOEAs; secondly, the first generation MOEAs are analyzed, which are characterized by simplicity, such as NSGA, NPGA, MOGA and so on, and the achievements and shortage during the first generation are also discussed. Thirdly, the MOEAs developed during the second generation, including SPEA, PAES,NSGA Ⅱ,NPGA2,PESA,Micro-GA and etc. , are detailed and compared, which use elitism to improve the efficiency. At last, some important research areas of MOEA are addressed.
出处
《计算机科学》
CSCD
北大核心
2007年第7期187-192,共6页
Computer Science
基金
国家自然科学基金(69975010
60374054)
山东省自然科学基金(Y2003G01)的支持
