DocumentCode :
2652626
Title :
Count Data Clustering Using Unsupervised Localized Feature Selection and Outliers Rejection
Author :
Bouguila, Nizar
Author_Institution :
Inst. for Inf. Syst. Eng., Concordia Univ., Montreal, QC, Canada
fYear :
2011
fDate :
7-9 Nov. 2011
Firstpage :
1020
Lastpage :
1027
Abstract :
This paper presents an unsupervised statistical model for simultaneous clustering, feature selection and outlier rejection in the case of count data. The proposed model is based on a finite discrete mixture to which a uniform component is added to ensure robustness to outliers and noise. The consideration of a finite mixture model is justified by its flexibility, its solid grounding in the theory of statistics and its competitive results. We derive a complete maximum a posteriori learning approach that does not require a priori knowledge about the number of outliers and the number of clusters. A rigorous expectation maximization (EM) algorithm, based on the formulation of a maximum a posteriori (MAP) estimation, is also provided. We report experimental results of applying our model to the challenging problems of visual scenes categorization and texture discrimination.
Keywords :
expectation-maximisation algorithm; feature extraction; maximum likelihood estimation; pattern clustering; unsupervised learning; data clustering; expectation-maximization algorithm; finite discrete mixture model; maximum a posteriori learning approach; outlier rejection; robustness; texture discrimination; unsupervised localized feature selection; unsupervised statistical model; visual scene categorization; Clustering algorithms; Data models; Feature extraction; Mathematical model; Vectors; Visualization; Mixture models; clustering; count data; feature selection; images categorization; outliers; texture;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Tools with Artificial Intelligence (ICTAI), 2011 23rd IEEE International Conference on
Conference_Location :
Boca Raton, FL
ISSN :
1082-3409
Print_ISBN :
978-1-4577-2068-0
Electronic_ISBN :
1082-3409
Type :
conf
DOI :
10.1109/ICTAI.2011.174
Filename :
6103465
Link To Document :
بازگشت