摘要
文 [1]指出 ,具有比例复制和自适应交叉、变异操作的遗传算法 (简称AGA)满足最优保存GA(简称EGA)的条件 ,则由EGA全局收敛的结论得出AGA也是全局收敛的 ;同时认为 ,AGA构成的Markov链为非时齐的 .本文给出了EGA的严格定义 ,指出了EGA全局收敛的本质 ,说明AGA实际并不属于EGA ,因此也不能沿用EGA全局收敛的结论 .在此基础上证明了AGA不能全局收敛 .最后仔细分析了AGA的遗传操作 ,说明AGA可由时齐Markov链来描述 .
Paper [1]points out that GA with proportional reproduction,adaptive crossover and mutation probability(AGA)meets the condition of elitist preserved GA(EGA)and concludes AGAs global convergence from EGAs global convergence conclusion.At the same time,its considered the Markov chain AGA generates is inhomogeneous. More normative definition of EGA is given and the essence of EGA global convergence is indicated.It illuminates AGA isnt one kind of EGA and its convergence analysis cannt follow the conclusion of EGAs.On the basis of it,AGAs inability to converge globally is proved.Finally,the genetic operation of AGA is analyzed carefully.It shows AGA can be described as a homogeneous Markov chain.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2001年第1期142-145,共4页
Control Theory & Applications
关键词
遗传算法
全局收敛性
计算效率分析
时齐性
genetic algorithm
adaptive crossover and mutation probability
convergence
homogeneity