基于进化算法的贝叶斯网络度量

利用图形模型描述进化过程中的种群,能够充分体现遗传信息的整体结构和内在联系,快速有效地进行网络图的构造与度量是将图形模型成功应用于进化算法的关键.贝叶斯网络是一类重要的图形模型,本文通过分析其结构,给出了用于刻划其度量的有关定义.对于染色体的某两个基因位置a、b,对应于贝叶斯网络图中的任意两个孤立的顶点,主要研究了a、b无连接、存在从a到b的连接以及存在从b到a的连接等三种基本图形的度量值关系.

METRIC IN BAYESIAN NETWORKS BASED ON THE EVOLUTIONARY ALGORITHM

Characterization of the population in evolutionary algorithms using graphical models is of advantage to express the whole construction and inherent relations of genetic information. The technique of speedy and effective to creation of a graph and its metric is the key of success to integrate these graphical models into an evolutionary algorithm. Bayesian networks are directed acyclic graphs, and they are one of the important graphical models. To analyze the structure of Bayesian networks based on evolutionary algorithm, some definitions with the property of the Bayesian networks metric were introduced in this paper. For any two gene locations a and b of chromosomes (that is, any two isolated vertices of Bayesian networks),he paper analyzed the metric relationship among three graphs containing an edge from a to b , from b to a and with no edge. At the same time the efficient approach to study the structure of population on evolutionary algorithm was also put forward.