Segmentation using EM algorithm
- 최초 등록일
- 2011.09.22
- 최종 저작일
- 2011.06
- 16페이지/ MS 워드
- 가격 10,000원
소개글
Pattern recognition(classification)과목을 통해서 수행한 과제로써
EM algorithm for gaussian mixture를 이용한 data segmentation을 한 report및 matlab 코드(보고서안에 포함)
목차
Contents
Background 2
Mixture Model (Gaussian) 2
Maximum likelihood on Mixture of Gaussians 3
EM for Gaussian Mixtures 5
Implementation 6
Condition & Assumption 6
Experiment Result 7
Conclusion 10
Programming codes 11
References 16
본문내용
1. Condition & Assumption
1) Sampling method: I generated three sets of 2-dimensional data randomly according to fixed means and covariance matrices. And each set is assumed to contain mixture of 3 Gaussian samples. The number of samples in the each set are 99 and the number of samples in the each clusters are 33.
2) Initializing prior: I generated k means and covariance matrices for the first prior values. In case of mean, I generated mean values randomly by the rand( ) function in matlab and the range of mean is [0 10]. And I fixed initial covariance matrices to [1 0; 0 1].
3) Stop condition for iteration: I fixed the threshold to “0.00005”. The iteration stops at the difference between previous log-likelihood and current log-likelihood is less than above threshold.
/
/
//
참고 자료
[1] R. Duta and P. Hart and D. Stork, Pattern classification, 2nd edition A Wiley-Interscience Publication, 2002.
[2] 패턴인식 개론 - MATLAB 실습을 통한 입체적 학습 개정판. 2009년 08월 22일 출간. 출판사 : 한빛미디어. 저자: 한학용. Page 210~211.