遗传算法的收敛性研究  

On Convergence of Genetic Algorithm

作　　者：屠昂燕[1] 陈建成[2] 

机构地区：[1]绍兴文理学院计算机系,绍兴312000 [2]浙江工业职业技术学院计算机系,绍兴312000

出　　处：《湖北第二师范学院学报》2008年第2期49-51,共3页Journal of Hubei University of Education

摘　　要：通过马尔可夫链方法,分析种群在解空间上概率分布情况以及收敛到最优解的概率,证明经典GA是不会收敛到最优解的,若在GA中保留每一代的最佳个体,则可以收敛到最优解。讨论全局收敛和过早收敛的原因,最后提出GA操作中应遵循的原则是改进GA搜索性能的关键。Based on the methods of Markov chain, the probability distribution of populations over the solution space and the probability of converging to the optimal solution were analyzed. It proves that classic GA is not converge to the optimal solution until the GA retains the best individuals of every generation. After that, the reasons of global convergence and premature convergence are discussed. Finally that GA operations should be guided by the principles, which is the key of improving the search performance of GA, is proposed.

关 键 词：遗传算法 收敛性 马尔可夫链 早熟收敛现象 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]