Presentation 2007-06-29
Semi-supervised Learning based on Dirichlet Process Mixture Models
Naonori UEDA, Takeshi YAMADA, Shuhei KUWATA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Focusing on the classifier design problem by using both labeled and unlabeled samples, we presents a new semi-supervised learning method based on the Dirichlet process mixture models. We employ the tied mixture model in which a feature vector space is shared among the classifiers for all classes so that both labeled and unlabed samples can be trained simultaneously. By introducing the Dirichlet process into the generation process of each component, the proposed method has a great advantage over the conventional tied mixture models in the sense that the proposed model can adaptively determine the appropriate number of components depending on a given set of trainig samples.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) semi-supervised learning / Dirichlet process mixture / tied mixture models / classifier design
Paper # DE2007-16,PRMU2007-42
Date of Issue

Conference Information
Committee DE
Conference Date 2007/6/21(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Data Engineering (DE)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Semi-supervised Learning based on Dirichlet Process Mixture Models
Sub Title (in English)
Keyword(1) semi-supervised learning
Keyword(2) Dirichlet process mixture
Keyword(3) tied mixture models
Keyword(4) classifier design
1st Author's Name Naonori UEDA
1st Author's Affiliation NTT Comunication Science Laboratories, NTT Corporation()
2nd Author's Name Takeshi YAMADA
2nd Author's Affiliation NTT Comunication Science Laboratories, NTT Corporation
3rd Author's Name Shuhei KUWATA
3rd Author's Affiliation NTT Comunication Science Laboratories, NTT Corporation
Date 2007-06-29
Paper # DE2007-16,PRMU2007-42
Volume (vol) vol.107
Number (no) 114
Page pp.pp.-
#Pages 6
Date of Issue