隐状态个数未知的隐马尔可夫0-1分布的贝叶斯推断  

Bayesian Inference for Implicit Markov 0-1 Distribution With Unknown Number of Hidden States

作　　者：李勇 刘鹤飞 Li Yong;Liu ttefei(College of Information Engineering;College of Mathematics and Statistics,Qujing Normal University,Qujing Yunnan 655011,China)

机构地区：[1]曲靖师范学院信息工程学院,云南曲靖655011 [2]曲靖师范学院数学与统计学院,云南曲靖655011

出　　处：《统计与决策》2018年第20期22-25,共4页Statistics & Decision

基　　金：云南省教育厅科研项目(2015C087Y);曲靖师范学院科研项目(2018QN004)

摘　　要：隐马尔可夫模型是一种基于随机过程的概率统计模型。文章研究隐状态个数未知时的隐马尔可夫0-1分布的贝叶斯推断。首先利用可逆跳跃马尔科夫链蒙特卡洛方法在可变维参数空间中抽样，根据隐状态个数的后验概率估计未知的隐状态个数。然后利用MCMC算法对隐状态个数确定时的隐马尔可夫0-1分布的参数进行估计。The Hidden Markov model is a kind of probability and statistics model based on stochastic process. This paper studies the Bayesian inference of the implicit Markov 0-1 distribution when the number of hidden states is unknown. Firstly, the paper uses the reversible jump Markov chain Monte Carlo method to sample the variable dimension parameter space, and then es- timates the number of unknown hidden states according to the posterior probability of the number of hidden states. Finally the pa- per uses the MCMC algorithm to estimate the parameters of the hidden Markov 0-1 distribution when the number of hidden states is determined.

关 键 词：隐马尔可夫模型 先验分布 可逆跳跃MCMC 贝叶斯推断 

分 类 号：O212.8[理学—概率论与数理统计]