基于贝叶斯网络理论的道德图学习  

Learning moral graphs based on Bayesian networks theory

作　　者：张剑飞[1] 王辉[2] 王双成[3] 

机构地区：[1]齐齐哈尔大学计算机与控制工程学院,黑龙江齐齐哈尔161006 [2]中央民族大学信息工程学院,北京100081 [3]上海立信会计学院信息科学系,上海201600

出　　处：《计算机应用研究》2009年第11期4042-4044,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(60675036)

摘　　要：贝叶斯网络的道德图是一种马尔可夫网络,是进行随机变量之间依赖关系分析、推理及预测的有力工具。基于道德图和贝叶斯网络间的密切联系,提出了一种基于贝叶斯网络理论进行道德图学习的方法。实验表明该方法能够显著提高道德图学习效率和可靠性,适合于多变量稀疏道德图学习。The moral graph of Bayesian network is a kind of Markov network. It is a powerful tool in analyzing dependent relationship, reasoning and predicting between variables. Bayesian networks have close relationship with moral graphs, proposed the constructing methods of moral graphs based on the Bayesian networks theory, the experiment indicates this method can obviously raise the moral graphs study efficiency and the reliability, the method is also fit to build multivariable sparse moral graphs.

关 键 词：贝叶斯网络 道德图 马尔可夫网络 非及物依赖 诱发依赖 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]