一种发现Bayesian网络结构的算法.pdfVIP

2005年中国模糊逻辑与计算智能联合学术会议论文集 859 一种发现Bayesian网络结构的算法 李小琳1，何湘东2，苑森淼1 长春130012) 摘要：Bayesian网络(贝叶斯网络)是一种进行不确定性推理的有力工具，它是一个有向无环图。本文提出了一种通 撞识别以及基于遗传算法和MDL(最小描述长度原理)标准的搜索打分方法确定边方向后得到相应的Bayesian网。 实验结果表明：该算法能有效的学习到Bayesian网络。 关键iS-Bayesian网；Bayesian网络骨架；碰撞识别；遗传算法 AMethodfor NetworksStructure DiscoveringBayesian 2 LIXiao—linl．HE YUANSen—mia01 Xiang—dong of Scienceand (1collegeComputer Technology，JilinUniversity,Changchun130012，Email：lixLl26@126．corn； 2VASOf China Vanda 130012，China) OperationGroup，Changchun networkisa Canbeusedtodouncertain networkisadirected Abstract：Bayesianpowerfultool，which reasoning．Bayesian methodof networkis methodlearns of theskeletona Network． acyclicgraph．A learningBayesian given．Firstly,this Bayesian Theskeletonofa Networkisanundirected ridoftheorientationofa bygetting Bayesian graph Bayesiannetwork．By basic between skeletonofa Networkis the variables，the built．Afterskeletonofa analyzingrelationship Bayesian Bayesian Networkis areorientedcollideridentificationand methodbasedon and found．edges by search＆scoring geneticalgorithm MDLmetric．Thena network Canbe show structure resultsthatthis Can correspondingBayesian built．Experimental algorithm learn network． effectivelyBayesian