基于边界的Markov网的发现  

Learning Markov Network Based on the Boundary

作　　者：何盈捷[1] 刘惟一[1] 

机构地区：[1]云南大学计算机科学系,昆明650091

出　　处：《计算机科学》2001年第9期78-82,65,共6页Computer Science

基　　金：国家自然科学基金(项目纪号69763003)

摘　　要：Markov network ts an another powerful tool besides Bayesian network which can be used to do uncertain inference. A method of learning Markov network automaticly from mass data based on boundary has been discussed in this paper. Taking advantage of an important conclusion in information theory ,we present an efficient boundary based Markov network learning algorithm. This algorithm only demands O(N2) times CI (conditional independence) test. We prove if the joint probability is strictly positive,then the found Markov network must be the minimal I_map of the sample.Markov network is an another powerful tool besides Bayesian network which can be used to do uncertain inference. A method of learning Markov network automaticly from mass data based on boundary has been discussed in this paper. Taking advantage of an important conclusion in information theory,we present an efficient boundary based Markov network learning algorithm. This algorithm only demands O(N2) times CI(conditional independence) test. We prove if the joint probability is strictly positive,then the found Markov network must be the minimal I_map of the sample.

关 键 词：MARKOV网 学习算法 不确定性推理 边界 人工智能 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]