马尔可夫网络的因子搜索算法  

Search Algorithm About Maximal Clique in Markov Network

作　　者：蒋文超[1] 许成[1] 曲敏[1] 

机构地区：[1]青岛大学数学与统计学院,青岛266071

出　　处：《青岛大学学报（自然科学版）》2016年第4期4-6,共3页Journal of Qingdao University(Natural Science Edition)

摘　　要：马尔可夫网络中,常用极大团定义因子,利用吉布斯分布把网络中包含的联合密度分布表示出来,以此对网络进行相应的概率解释,其中搜索极大团这一过程变得尤为重要。因此在马尔可夫网络的理论基础上,结合图论知识,提出一个马尔可夫网络中的极大团搜索算法。搜索算法可以对输入的网络进行缩减,简化网络,并能得到网络中的所有极大团,同时算法的运算结果具有唯一性。In the Markov networks, the maximal clique are often defined as factors, and Gibbs distribution was used to denote the joint density distribution of the network for the probability interpretation of the network. The process of finding maximal clique becomes very important. The paper proposes a searching algorithm of maximal clique. It＇s based on the theory of Markov networks, and considering of the graph- ical theory. The algorithm can reduce and simplify the network, at last it can find all the maximal cliques of the graph. Meanwhile, the algorithm has the unique results.

关 键 词：马尔可夫网络 无向图 极大团 因子化 概率图模型 

分 类 号：O221.9[理学—运筹学与控制论]