摘要
提出了针对个体变异率和种群数目的 2种自适应方法 .算法中个体变异率根据其适度值在种群中的排序自适应调整 ,使优良个体具有较小的变异率继续进化 ,而使种群中较差个体具有较大变异率 ,增强了种群搜索能力 .同时根据种群个体适度值方差动态调整变异率曲线 ,种群数目调整则根据最优个体更新率动态增大 ,以动态适应解空间的规模避免采样误差造成的进化停滞 .通过在不同尺度的NKLandscape上与传统的简单遗传算法 (SGA)比较可得 。
Two parameter adaptation methods are presented for genetic algorithm. Mutation probability is assigned to each individual according to its sort order of fitness in the population. Individuals with above average fitness have lower mutation probabilities and continually evolve to better ones, while less fit individuals are assigned with higher mutation probabilities to search wider area. Meanwhile, the populations fitness variance is used to adjust the probability curve. Population size is doubled when no best individual is updated after c ertain numbers of generations. Experiments are carried out by comparing multi scal e NK Landscapes with simple genetic algorithm (SGA). Results show that the optimization ability of genetic algorithm is improved remarkably by introducing the presente d parameter adaptation methods.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2004年第4期553-556,共4页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目 ( 60 2 75 0 41)
南瑞继保研究生论文基金资助项目 ( 2 0 0 3 )
关键词
遗传算法
变异率
种群数
自适应
genetic algorithm
mutation probability
population size
adaptation methods
