Presentation 2005/3/21
A variational Bayes inference for a mixture of von Mises-Fisher distribution
Akihiro TANABE, Kenji FUKUMIZU, Shigeyuki OBA, Shin ISHII,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) To deal with high dimensional directional data is getting important in various applications like analyses of text and gene expression data. In this report, we propose variational Bayes (VB) algorithm for mixture of von Mises-Fisher distributions, which is a probability density function defined on a unit hypersphere. We compare the predictive distribution obtainded by the VB algorithm with that by the maximum likelihood estimation, in order to check the effectiveness of this algorithm. Besides, we introduce an approximation method of the logarithm of the first kind modified Bessel function.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) von Mises-Fisher distribution / variational Bayes / free energy / predictive distribution / first kind modified Bessel function
Paper # NC2004-159
Date of Issue

Conference Information
Committee NC
Conference Date 2005/3/21(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 Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A variational Bayes inference for a mixture of von Mises-Fisher distribution
Sub Title (in English)
Keyword(1) von Mises-Fisher distribution
Keyword(2) variational Bayes
Keyword(3) free energy
Keyword(4) predictive distribution
Keyword(5) first kind modified Bessel function
1st Author's Name Akihiro TANABE
1st Author's Affiliation Nara Institute of Science and technology()
2nd Author's Name Kenji FUKUMIZU
2nd Author's Affiliation The Institute of Statistical Mathematics
3rd Author's Name Shigeyuki OBA
3rd Author's Affiliation Nara Institute of Science and technology
4th Author's Name Shin ISHII
4th Author's Affiliation Nara Institute of Science and technology
Date 2005/3/21
Paper # NC2004-159
Volume (vol) vol.104
Number (no) 758
Page pp.pp.-
#Pages 6
Date of Issue