计算机工程 ›› 2011, Vol. 37 ›› Issue (10): 143-145.doi: 10.3969/j.issn.1000-3428.2011.10.048

• 人工智能及识别技术 • 上一篇    下一篇

基于进化MCMC的DBN学习算法

郭 鹏 1,2,李乃祥 2,刘同海 2   

  1. (1. 天津大学计算机科学与技术学院,天津 300072;2. 天津农学院计算机科学与信息工程系,天津 300384)
  • 出版日期:2011-05-20 发布日期:2011-05-20
  • 作者简介:郭 鹏(1974-),男,讲师、博士研究生,主研方向:进化计算,取样统计学习;李乃祥,教授;刘同海,讲师
  • 基金项目:
    中华人民共和国环境保护部公益基金资助项目(HB2009 ZRBHQ-03);中华人民共和国野外科学观测研究站基金资助项目(NGEFS-09-04)

DBN Learning Algorithm Based on Evolutionary MCMC

GUO Peng 1,2, LI Nai-xiang 2, LIU Tong-hai 2   

  1. (1. School of Computer Science and Technology, Tianjin University, Tianjin 300072, China; 2. Department of Computer Science and Information Engineering, Tianjin Agricultural University, Tianjin 300384, China)
  • Online:2011-05-20 Published:2011-05-20

摘要: 提出利用进化MCMC算法进行动态贝叶斯网络(DBN)学习的方法。在数据缺省情况下利用EM算法进行贝叶斯网络参数学习,结构学习部分生成多条备选的贝叶斯网络染色体,对染色体进行变异操作和交叉操作,在遗传操作中根据温度参数和贝叶斯网络及贝叶斯信息准则来构造MCMC函数,并利用MCMC函数进行贝叶斯网络学习。每一代进化后,将贝叶斯信息评分最大的贝叶斯网络作为结构学习的结果。实验结果验证了该方法性能的稳定性。

关键词: 动态贝叶斯网络, EM算法, 贝叶斯信息准则, 进化MCMC

Abstract: Dynamic Bayesian Network(DBN) learning with evolutionary MCMC algorithm is presented. Parameter learning with absent data is done with EM algorithm. In structure learning, multiple Bayesian network chromosomes are generated as candidates and are processed with mutation and crossover. The structures are learned with MCMC function which is obtained with temperatures and Bayesian Information Criterion(BIC) scores of corresponding Bayesian networks. In each generation, the Bayesian network with the maximal BIC score is selected as the result of the structure learning. Experimental results proves the stability of the method’s performance.

Key words: Dynamic Bayesian Network(DBN), EM algorithm, Bayesian information criterion, evolutionary MCMC

中图分类号: