基于t分布变异的进化规划  

Evolutionary Programming Using Mutations Based on the t Probability Distribution

作　　者：周方俊[1] 王向军[1] 张民[1] 

机构地区：[1]海军工程大学电气与信息工程学院,湖北武汉430033

出　　处：《电子学报》2008年第4期667-671,共5页Acta Electronica Sinica

摘　　要：参考已提出的各种变异算子,提出了基于t分布变异的进化规划.在该算法中,使用t分布将高斯分布和柯西分布衔接起来,改变自由度n可以调整其变异性能.通过仿真得到了t算子随n和σ变化的一般规律,初步总结了进化代数、变异尺度和求解精度三者之间的关系.在此基础上,给出了一个基于t算子的自适应性进化规划算法,该算法将高斯变异和柯西变异的优点结合起来,仿真表明算法具有良好的性能.Consulted some previous mutations,this paper proposed an evolutionary programming（EP）using mutations based on the t probability distribution（tEP）. The t probability distribution can connect with Gaussian and Cauchy probability distribution.. Its variance is changeable by adjusting degree of freedom n.Rules of t mutations according to the change ofn and standard deviation are analyzed by simulations, and the relationship among number of generations, standard deviation and minimum fitness are briefly summarized. Then an EP using adaptive t mutations（AtEP）is put forward at the end. The AtEP uses several t mutations with different n at the same time, so it can integrate the advantages of Gaussian and Cauchy mutation. And simulation evidence shows the algorithm is better than classical EP algorithm.

关 键 词：进化规划 算子 T分布 T算子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]