匹配于进化种群的树形贝叶斯网络

Learning tree-like Bayesian networks from the evolutionary population

为了构造匹配于进化种群的树形模型，首先研究了基于贝叶斯狄利克雷度量构建树形网络结构的方法，得出关键在于搜索每一个节点的最大值父节点。然后提出了节点的势及对称节点概念，证得节点的势与网络结构的连接方向关系密切，以及节点与其对称节点在贝叶斯网络图中具有相同的度量属性。最后给出了仿真分析结果，进一步表明本文提出的方法能够依据数据信息搜索到具有最大度量值的树形网络结构。

To learn tree-like Bayesian networks from a data set to match evolutionary population, firstly, an approach to construct a tree model based on Bayesian-Dirichlet metric is developed by searching parent node with the highest score. Secondly, the definitions of the potential of node and the symmetric node are given, and it is derived that the potential of nodes is very important for arc direction of network structures. It is also showed that node and its symmetric node have the same Bayesian-Dirichlet metric score in the network structures. Finally, the simulation and analysis results show that the approach is efficient and reliable, and it provides a new method and theoretical supports for creating tree-like netowrk structure with the highest score.