计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (30): 43-46.DOI: 10.3778/j.issn.1002-8331.2008.30.013

• 理论研究 • 上一篇    下一篇

期望最大算法及其应用(2)

李昌利,沈玉利

  

  1. 广东海洋大学 信息学院,广东 湛江 524088
  • 收稿日期:2007-11-20 修回日期:2008-01-14 出版日期:2008-10-21 发布日期:2008-10-21
  • 通讯作者: 李昌利

Tutorial of EM algorithm and its application:part Ⅱ

LI Chang-li,SHEN Yu-li   

  1. School of Information Engineering,Guangdong Ocean University,Zhanjiang,Guangdong 524088,China
  • Received:2007-11-20 Revised:2008-01-14 Online:2008-10-21 Published:2008-10-21
  • Contact: LI Chang-li

摘要: EM算法是实现极大似然估计的一种有效方法,主要用于非完全数据的参数估计。文章的第一部分已经详细介绍了算法的基本原理,这部分内容着重介绍算法的各种应用,特别是高斯混合模型、隐马尔科夫模型和因子分析中的参数估计。

关键词: 期望最大(EM), 高斯混合模型, 隐马尔科夫模型, 因子分析

Abstract: EM algorithm is an effective method for Maximum-Likelihood Estimate(MLE),which is mainly used to estimate parameters of incomplete data.While part one of this tutorial is mainly on the basic principle of EM algorithm,this part deals with its applications especially in parameters estimation for mixtures of Gaussians,hidden Markov model and factor analysis.

Key words: Expectation-Maximization(EM), Mixtures of Gaussians(MoG), Hidden Markov Model(HMM), Factor Analysis(FA)