Presentation 1999/1/21
Probabilistic Mixture Subspace Method : Pattern Recognition by Mixture of Factor Analyzers
Naonori UEDA, Ryohei NAKANO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We present a classification method using mextures of factor analyzers. In this method class probability for each class is estimated from given training data in the same class by approximating the feature space with the mixture of affine subspaces with less dimensionality. The proposed method is similar to the conventional subspace methods in the sense that the classification is done based on the dimensionality reduction of the feature space. However, it is essentially different from them in that since it is formulated as a probabilistic model, it can compute posterior probabilities for unknown data. That is, the proposed method can perform Bayes classification. We apply the proposed method to real data sets and show that it outperforms the conventional subspace method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Pattern recognition / Mixtures of factor analyzers / Subspace method / Bayes classification
Paper # NC98-72
Date of Issue

Conference Information
Committee NC
Conference Date 1999/1/21(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Vice Chair

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Probabilistic Mixture Subspace Method : Pattern Recognition by Mixture of Factor Analyzers
Sub Title (in English)
Keyword(1) Pattern recognition
Keyword(2) Mixtures of factor analyzers
Keyword(3) Subspace method
Keyword(4) Bayes classification
1st Author's Name Naonori UEDA
1st Author's Affiliation NTT Communication Science Laboratories()
2nd Author's Name Ryohei NAKANO
2nd Author's Affiliation NTT Communication Science Laboratories
Date 1999/1/21
Paper # NC98-72
Volume (vol) vol.98
Number (no) 526
Page pp.pp.-
#Pages 7
Date of Issue