一种隐式法的贝叶斯网络结构学习  

Learning Bayesian Network Structure based on Implicit Method

作　　者：何宗耀[1] 王刚[1] 

机构地区：[1]河南城建学院计算机科学与工程系,河南平顶山467044

出　　处：《河南师范大学学报（自然科学版）》2011年第6期150-153,共4页Journal of Henan Normal University(Natural Science Edition)

基　　金：河南省科技攻关重点项目(092102210251)

摘　　要：通过分析K2,BIC,AIC和IM等方式的原理,改进K2算法,在不考虑先验知识的基础上,创建了新的基于隐式网络的打分函数取代了原有算法的评分规则,实现贝叶斯网络结构学习.仿真实验结果表明,针对标准数据集学习,隐式法的贝叶斯网络学习算法在没有先验知识的条件下和依赖先验知识的基于BDe评分的K2算法相比收敛速度和准确率有一定的改进.In order to overcome the problem that most of the priori knowledge for learning Bayesian network structure is difficulty to obtain, implicit distribution is utilized to improve K2 algorithm. An implicit method score based on implicit network is constructed to replace the score of original algorithm. Experiment results used in standard data sets show that this method without priori knowledge can effectively learn a comparative correct and steady Bayesian network structure, which is as good as the K2 algorithm based on BDe score.

关 键 词：隐式分布 贝叶斯网络 结构学习 隐式法 K2 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]