基于EM算法极大似然参数估计探讨.pdfVIP

32 4 () Vol.32　No.4 2002 12 Journal of Henan University (Natural Science) Dec.2002 EM 1 1 2 孙大飞 , 陈志国, 刘文举 (1.河南大学计算机与信息工程学院, 河南 开封475001;2.中国科学院自动化所模式识别国家重点实验 , 北京100080) 　:EM , EM , . :;;;;EM :　　　:A :1003-4978(2002)04-0035-07 Discussion of Maximum Likelihood Parameter Estimation Based On EM Algorithm 1 1 2 SUN Da-fei , CHEN Zhi-guo , IU Wen-ju (1.Institute of Comp uter and I nf ormation Engineering, Henan University, K aif eng 475001, Henan China ; 2 .National Laboratory of Pattern Recognition Institute of Automation CA , Beij ing 100080, China) Abstract :In this paper, we introduce the essence of the algorithm, study the method of mixture densities parameter estimation based on EM Algorithm, and give a computer emulation example to illustrate convergence and validity of EM algorithm. Key words :parameter estimation ;likelihood function ;maximum likelihood parameter estimation ;incomplete data likelihood function ;EM algorithm 0 　 [ 1, 2, 4～9] , , , 、、., , . , , , , , EM [4～6] .EM , , , ., EM . 1　 , . x , p (x |θ), θ={θ1, …θM }, θ. N x , …x , p (x |θ).X ={x , …x }, 1 N 1 N 　:2002-09-30 　:、(0312001900)、 (2002KYCX007) 　:(1976-), , . 　　　36 (), 2002 , 32 4 N p (X |θ)= p (x i |θ) L (θ|X ). (1.1) ∏ i=1 L (θ|X ).X , L (θ|X )θ.L (θ |X )θ, θ =arg maxL (θ|X ),