【摘要】 进化算法是研究全局优化算法中最重要的随机算法之一,本文给出了进化规划和进化策略的变异算子的数学描述,并提出变异函数的概念,在此基础上,给出了用均匀分布的随机数构造变异算子的几种方法和若干例子.结果表明,利用本文给出的方法,不仅可以构造出目前进化策略和进化规划算法普遍采用的几种变异算子,还可以构造出新的变异算子.针对一般的变异算子,在不要求目标函数连续的情况下,证明了保持最优个体的进化规划和进化策略,迭代产生的最优个体的函数值收敛到问题的最优值的ε-邻域的概率为1.更多还原

【Abstract】 In this paper,a mathematical description of mutation operator and a concept of mutation function for the study of evolutionary programming and evolution strategies are proposed,and several methods and examples of constructing mutation op- erator with uniformly distributed random variables are presented.Based on these new results not only the classical mutation operators of EP and ES can be obtained,but also new mutation operators can be easily constructed out.Further,under weakened conditions(without the assumption of continuity of objective function),the convergence theorem of EP and ES with the general mutation operators is also given.更多还原