分布估计算法中适应度函数模型检验及应用  

Validity Testing to Fitness Function Modelling in Estimation of Distribution Algorithm and Application

作　　者：吴养会[1] 卢恩双[1] 邓业胜[1] 秦宝福[2] 

机构地区：[1]西北农林科技大学理学院应用数学系,陕西杨凌712100 [2]西北农林科技大学生命学院,陕西杨凌712100

出　　处：《数学的实践与认识》2008年第18期98-104,共7页Mathematics in Practice and Theory

基　　金：中央环保专项子项目(14111103);西北农林科技大学人才基金资助

摘　　要：在一种基于马尔可夫网络的分布估计算法中,利用解的适应度函数模型表示变量的概率分布.通过相关系数检验建立了一种衡量适应度函数模型有效性的方法,同时该方法也可以用来确定初始种群规模大小.在二维相关变量结构下,将该方法应用于一种多变量生物动态模型的优化问题,结论表明通过该方法可以选择适当的种群规模,保证适应度函数模型的有效性,并提高算法的优化性能.DEUM is one of the novel Estimation of Distributions based on Markov Network. In DEUM, fitness function modelling was built to express the probabilistic distribution of variables. In this paper, we set up a method to test the validity of fitness function modelling by standard statistical correlation coefficient, during the process of getting the correlation, it can easily get the size of initial population. We apply this approach to a multivariate dynamic Biocontrol problem in term of Chain structure. Experimental results indicate that the fitness function modelling is valid and capable of getting more accurately optimal solutions.

关 键 词：分布估计算法 马尔可夫结构 适应度函数 生物控制模型 

分 类 号：O211.62[理学—概率论与数理统计]