An Enhanced em algorithm using maximum entropy distribution as initial condition

Guorong Xuan; Yun-Qing Shi; Peiqi Chai; Patchara Sutthiwan

An Enhanced em algorithm using maximum entropy distribution as initial condition

Guorong Xuan, Yun-Qing Shi, Peiqi Chai, Patchara Sutthiwan

Electrical and Computer Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

4 Scopus citations

Abstract

The conventional EM algorithms may suffer from the following two problems. First, it may converge to a local maximum. Second, the algorithm may suffer from singularity. A novel Enhanced EM algorithm (EEM) using a realization of maximum-entropy uniform distribution as initial condition is proposed. A global optimal solution can be obtained. In addition, a positive perturbation scheme is adopted to avoid singularity. Experimental results have demonstrated that the EEM is simple and effective compared with some prior arts.

Original language	English (US)
Title of host publication	ICPR 2012 - 21st International Conference on Pattern Recognition
Pages	849-852
Number of pages	4
State	Published - Dec 1 2012
Event	21st International Conference on Pattern Recognition, ICPR 2012 - Tsukuba, Japan Duration: Nov 11 2012 → Nov 15 2012

Other

Other	21st International Conference on Pattern Recognition, ICPR 2012
Country/Territory	Japan
City	Tsukuba
Period	11/11/12 → 11/15/12

All Science Journal Classification (ASJC) codes

Computer Vision and Pattern Recognition

Cite this

@inproceedings{ed260ed17086467484eb3938bf649ea1,

title = "An Enhanced em algorithm using maximum entropy distribution as initial condition",

abstract = "The conventional EM algorithms may suffer from the following two problems. First, it may converge to a local maximum. Second, the algorithm may suffer from singularity. A novel Enhanced EM algorithm (EEM) using a realization of maximum-entropy uniform distribution as initial condition is proposed. A global optimal solution can be obtained. In addition, a positive perturbation scheme is adopted to avoid singularity. Experimental results have demonstrated that the EEM is simple and effective compared with some prior arts.",

author = "Guorong Xuan and Yun-Qing Shi and Peiqi Chai and Patchara Sutthiwan",

year = "2012",

month = dec,

day = "1",

language = "English (US)",

isbn = "9784990644109",

pages = "849--852",

booktitle = "ICPR 2012 - 21st International Conference on Pattern Recognition",

note = "21st International Conference on Pattern Recognition, ICPR 2012 ; Conference date: 11-11-2012 Through 15-11-2012",

}

TY - GEN

T1 - An Enhanced em algorithm using maximum entropy distribution as initial condition

AU - Xuan, Guorong

AU - Shi, Yun-Qing

AU - Chai, Peiqi

AU - Sutthiwan, Patchara

PY - 2012/12/1

Y1 - 2012/12/1

N2 - The conventional EM algorithms may suffer from the following two problems. First, it may converge to a local maximum. Second, the algorithm may suffer from singularity. A novel Enhanced EM algorithm (EEM) using a realization of maximum-entropy uniform distribution as initial condition is proposed. A global optimal solution can be obtained. In addition, a positive perturbation scheme is adopted to avoid singularity. Experimental results have demonstrated that the EEM is simple and effective compared with some prior arts.

AB - The conventional EM algorithms may suffer from the following two problems. First, it may converge to a local maximum. Second, the algorithm may suffer from singularity. A novel Enhanced EM algorithm (EEM) using a realization of maximum-entropy uniform distribution as initial condition is proposed. A global optimal solution can be obtained. In addition, a positive perturbation scheme is adopted to avoid singularity. Experimental results have demonstrated that the EEM is simple and effective compared with some prior arts.

UR - http://www.scopus.com/inward/record.url?scp=84874564786&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84874564786&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84874564786

SN - 9784990644109

SP - 849

EP - 852

BT - ICPR 2012 - 21st International Conference on Pattern Recognition

T2 - 21st International Conference on Pattern Recognition, ICPR 2012

Y2 - 11 November 2012 through 15 November 2012

ER -

An Enhanced em algorithm using maximum entropy distribution as initial condition

Abstract

Other

All Science Journal Classification (ASJC) codes

Other files and links

Fingerprint

Cite this