遗传算法中初始种群与交叉、变异率对解的影响及其解决方案  

Effects of the Initial Population,Crossover and Mutation Rate to the Results of Genetic Algorithms and a Possible Solution Scheme

作　　者：唐世浩[1] 朱启疆[1] 

机构地区：[1]北京师范大学资环系,北京100875

出　　处：《科技通报》2001年第3期1-7,共7页Bulletin of Science and Technology

基　　金：国家自然科学基金资助项目!( 49871 0 55)

摘　　要：本文的研究表明 ,在相同的遗传算子下 ,初始种群性状和数量以及交叉、变异率的确定对算法收敛速度和结果的影响不能忽略 .初始种群或交叉、变异率选择不当 ,将增加迭代次数 ,甚至直接导致算法陷入局部最优解 .为此 ,本文提出一种基于空间分割的遗传算法及初始种群产生和种群数量确定方法 ,并根据有关文献 ,提出一种自适应交叉、变异率方法 .实际计算表明 ,该算法在很大程度上避免了算法收敛于局部最优点 。Our studies show that under the same genetic operators,effects of parameters,such as initial population,crossover rate and mutation rate,can't be ignored. Randomly generated initial population is not always suitable, and it's sometimes responsible for the unstable solutions. Unsuitable crossover and mutation rate can cause the same problem.For these reasons,an improved genetic algorism is presented in this paper.Self adjusting crossover and mutation rate are used and the initial population and population size are determined based on space division in this algorism.Results show that this method can avoid local convergence greatly.

关 键 词：遗传算法 初始种群 交叉 变异率 遗传算子 子空间分割 GA算法 

分 类 号：O242.23[理学—计算数学]