Presentation 2012-11-08
An Efficient Sampling Algorithm for Bayesian Variable Selection
Takamitsu ARAKI, Kazushi IKEDA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In Bayesian variable selection, a Gibbs variable selection (GVS) is one of the most famous sampling algorithms, and has been used in various models. The efficiency of the GVS strongly depends on parameters of a proposal distribution and pseudo-priors, and the GVS determines their parameters based on a pilot run for a full model. However the parameters shift the pseudo-priors from the marginal posterior distributions, and make the scale of the proposal distribution an improper value in many cases. In this paper, we propose an algorithm that adapts the parameters while it runs, and confirm that our algorithm is more efficient than the conventional GVS by a experiment of Bayesian variable selection of a logistic regression model.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Gibbs Variable Selection / Adaptive Markov Chain Monte Carlo / Bayesian logistic regression model
Paper # IBISML2012-75
Date of Issue

Conference Information
Committee IBISML
Conference Date 2012/10/31(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 Information-Based Induction Sciences and Machine Learning (IBISML)
Language ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) An Efficient Sampling Algorithm for Bayesian Variable Selection
Sub Title (in English)
Keyword(1) Gibbs Variable Selection
Keyword(2) Adaptive Markov Chain Monte Carlo
Keyword(3) Bayesian logistic regression model
1st Author's Name Takamitsu ARAKI
1st Author's Affiliation Graduate School of Information Science, Nara Institute of Science and Technology()
2nd Author's Name Kazushi IKEDA
2nd Author's Affiliation Graduate School of Information Science, Nara Institute of Science and Technology
Date 2012-11-08
Paper # IBISML2012-75
Volume (vol) vol.112
Number (no) 279
Page pp.pp.-
#Pages 5
Date of Issue