摘要
论文提出了一种新的遗传操作算子-随机多父辈适应度值加权交叉算子(RandomParent-numberFitness-WeightedCrossover,RPFWX)。RPFWX在2到群体规模之间随机决定参与交叉操作的父代数量,以克服遗传算法的性能对于多父辈数量的敏感性。RPFWX根据父代个体的适应度值赋予较优秀的父代个体以较大的影响因子。父代个体根据该影响因子来决定其对于子代个体贡献的大小。RPFWX是包括两父辈算术交叉、uniformcrossover以及多父辈频率扫描交叉算子(theOccurrenceBasedScanningCrossover)等在内的多种交叉算子的概括和推广,并通过其凸空间封闭特性来保证子代个体的合法性。最后通过实验结果验证了RPFWX算子的优越性。
A new crossover operator named Random Parent-number Fitness-Weighted Crossover(RPFWX)is introduced and studied in this paper.RPFWX randomly choose the parent number in the range from2to the population size since the performance of GA with Multi-Parent Recombination is sensitive to the parent number.RPFWX endues the fitter parent with a bigger influencing factor,and parents decide their contribution to offspring according to the factors.RPFWX is the generalization of the Arithmetical Crossover,Uniform Crossover,the Multi-parent Occurrence Based Scan-ning Crossover,and etc.RPFWX can guarantee the validity of offspring because it is a convex combination of all par-ents.Experiment result validates the advantages of RPFWX.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第6期7-10,46,共5页
Computer Engineering and Applications
基金
国家自然科学基金重点项目资助(编号:60234020)
关键词
遗传算法
多父辈交叉
交叉算子
Genetic algorithm,Multi-parent recombination,Crossover operator