Probability Graph Model And Its Independence Study

Probabilistic graph model show the relationship between the probabilistic distribut ion of variables by using graphical structure implied structure characteristics,not only makes the relationship more intuitive,but also simplifies the operation.Because of this advantage,probabilistic graph model is very important in uncertain reasoning,and has a very good performance in medical diagnosis,artificial intelligence,data mining and so on.This paper studies the representation of the probabilistic graph mode,and the rel evant knowledge of the independence,discusses the relationship between the distributio n and graph,and present several new algorithms.Firstly the paper introduces several fr equently-used probabilistic distributions,the CPD of the separated points such as table CPD and deterministic CPD,the CPD of the continuous points such as Gaussian mo del,as well as the mixed model.Part two introduces the Bayesian networks model repr esentation detailedly,including the network structure and the independence,in particular,this paper presents a bayesian networ-simplicity Bayesian networks,then it discusses the relationship between the distribution and the figure in the Bayesian network,and gives a algorithm to build the network under the condition of the distribution is giv en,and provides a thinking to find the order by using the parent node.Part three intro duces the Markov network structure and parameterization problem,and the parameteriz ation process is generally regarded as the factorization process.Two search algorithms are given in this section,and about the maximize clique and the maximum clique re spectively.They make preparations for the factorization.And then the paper introduces the independences in the Markov network,and illustrates the inclusion relationship bet ween them.Next the paper introduces the transform from Bayesian networks to Marko v networks.Finally,the article makes a summary,as well as explain the faced problem in the fields about the transformation of the distribution and the figure,the improveme nt of the algorithms and the transformation between the two network.