算术杂交算子的问题相关性  

Correlation between arithmetic crossovers and problems

作　　者：周永华[1] 谢杰华[1] 

机构地区：[1]广西大学电气工程学院,广西南宁530004

出　　处：《广西大学学报（自然科学版）》2011年第6期1000-1003,1015,共5页Journal of Guangxi University（Natural Science Edition）

基　　金：广西自然科学基金资助项目(桂科自0991060)

摘　　要：针对实数编码遗传算法中的算术杂交算子是否存在问题相关性进行了实验研究。首先,用几种常用的算术杂交算子对测试函数进行优化实验,结果表明,选用何种算术杂交算子算法更有效与具体优化问题有关,即算术杂交算子存在问题相关性,且它随变异概率的变化而呈现比较复杂的变化。其次,提出了根据算术杂交算子在无变异时的优化效果以及有变异时与变异算子的协同优化效果,挑选最适合某一具体问题的算术杂交算子的方法。这对应用实数编码遗传算法高效地求解函数优化问题是有帮助的。An experiment is investigated to find out whether there is a correlation between arithmetic crossovers of real coded genetic algorithms and corresponding problems.Firstly,several common used arithmetic crossovers are used in the experiments for a test function.The experimental results show that choosing which crossover to make the algorithm performing more effectively is related to the problem to be solved,that is there is the correlation between arithmetic crossovers and corresponding problems,and it appears to be changed complicatedly with the mutation probability.Secondly,the method of how to choose an appropriate arithmetic crossover for an optimization problem is proposed by investigating in numeric experiments the performance of each arithmetic crossover without mutation,and the synergetic performance of the combination of an arithmetic crossover and mutation.This is helpful for effectively using real coded genetic algorithms to solve function optimization problems.

关 键 词：遗传算法 算术杂交算子 相关性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]