摘要
多目标优化问题中,人们往往只是对目标空间的某一区域感兴趣,因此这就需要在这一特定的区域能够得到比较稠密的Pareto解,但传统的方法权值法无法满足这种需求而且不能处理目标空间是非凸的情况,遗传算法虽然是现在公认的处理多目标优化问题比较有效的方法,但遗传算法是在目标空间内进行全空间寻优,因此最终得到的Pareto解是均匀分布的,这样遗传算法也不能满足人们的这一要求。针对这个问题提出了基于偏好的多目标遗传算法,把个人偏好加到优化过程中,利用偏好信息来引导优化方向,通过仿真把该算法和权值法、NSGA-II进行比较,结果证明了该算法的可行性和有效性。
In multi-objective optimization problem,many people are only interested in a special part of the objective space,so which should has enough solutions.Weighted method cannot satisfy this demand and in the same time it cannot deal with nonconvex case; although genetic algorithm is an ideal method for multi-objective optimization; its solutions are uniformly distributed in the objective space,so it also cannot.Due to this question,we bring forward a new multi-objective genetic algorithm incorporated preference into optimization process to direct optimization,and compare this algorithm with weighted method and NSGA-Ⅱ through simulation,which validates this algorithm＇s feasibility and advantage.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第9期24-26,共3页
Computer Engineering and Applications
基金
国家自然科学基金(the National Natural Science Foundation of China under Grant No.60674070)
关键词
多目标优化
遗传算法
偏好
权值法
NSGA—Ⅱ
multi-objective optimization
Genetic Algorithm
preference
weighted method
NSGA-Ⅱ
