A note on the parameterized EM method

被引:3
作者
Roland, Christophe [1 ]
机构
[1] Univ Montpellier 2, Inst Math & Modelisat Montpellier, CNRS, UMR 5149,Equipe Probabil & Stat, F-34095 Montpellier 5, France
关键词
EM algorithm; Fixed point iteration; Acceleration; Convergence; Spectral radius; ACCELERATION;
D O I
10.1016/j.spl.2010.04.017
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
We study and analyze the parameterized method in view of the acceleration of the EM algorithm. Some theoretical results and discussions on the choice of the step size improve the understanding of the method and its limits. (C) 2010 Elsevier B.V. All rights reserved.
引用
收藏
页码:1354 / 1357
页数:4
相关论文
共 12 条
[1]  
[Anonymous], 2008, EM ALGORITHM EXTENSI
[2]  
Bauer Eric., 1997, Proceedings of the 13th Conference on Uncertainty in Artifical Intelligence, P3
[3]   Parabolic acceleration of the EM algorithm [J].
Berlinet, A. ;
Roland, C. .
STATISTICS AND COMPUTING, 2009, 19 (01) :35-47
[4]   MAXIMUM LIKELIHOOD FROM INCOMPLETE DATA VIA EM ALGORITHM [J].
DEMPSTER, AP ;
LAIRD, NM ;
RUBIN, DB .
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-METHODOLOGICAL, 1977, 39 (01) :1-38
[5]  
Hesterberg T, 2005, P AM STAT ASS STAT C, P2101
[6]   Acceleration of the EM algorithm by using quasi-Newton methods [J].
Jamshidian, M ;
Jennrich, RI .
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-METHODOLOGICAL, 1997, 59 (03) :569-587
[7]   CONJUGATE-GRADIENT ACCELERATION OF THE EM ALGORITHM [J].
JAMSHIDIAN, M ;
JENNRICH, RI .
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1993, 88 (421) :221-228
[8]  
LANGE K, 1995, STAT SINICA, V5, P1
[9]  
LOUIS TA, 1982, J ROY STAT SOC B MET, V44, P226
[10]  
Ortiz LE, 1999, UNCERTAINTY IN ARTIFICIAL INTELLIGENCE, PROCEEDINGS, P512