Presentation 2023-12-21
On the benefits of Partial Stochastic Bayesian Neural Networks
Koki Sato, Daniel Andrade,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Bayesian neural networks (BNNs) can model uncertainty in the prediction results better than ordinary neural networks. However, accurate approximation of the posterior distribution via Markov chain Monte Carlo (MCMC) methods is computationally expensive due to the large number of parameters in the BNNs. Therefore, in this study, we evaluate the effectiveness of fixing some of the model parameters using the Maximum-A-Posterior (MAP) method, and only sample from a small set of remaining parameters. Extensive evaluation of various regression datasets confirm the effectiveness of the partial stochastic BNNs.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Neural Networks / Bayesian Inference / Markov Chain Monte Carlo / sampling efficiency
Paper # IBISML2023-36
Date of Issue 2023-12-13 (IBISML)

Conference Information
Committee IBISML
Conference Date 2023/12/20(2days)
Place (in Japanese) (See Japanese page)
Place (in English) National Institute of Informatics
Topics (in Japanese) (See Japanese page)
Topics (in English) machine learning, etc.
Chair Masashi Sugiyama(Univ. of Tokyo)
Vice Chair Toshihiro Kamishima(AIST) / Koji Tsuda(Univ. of Tokyo)
Secretary Toshihiro Kamishima(NTT) / Koji Tsuda(Hokkaido Univ.)
Assistant Yoshinobu Kawahara(Osaka Univ.) / Taiji Suzuki(Univ.of Tokyo)

Paper Information
Registration To Technical Committee on Information-Based Induction Sciences and Machine Learning
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) On the benefits of Partial Stochastic Bayesian Neural Networks
Sub Title (in English)
Keyword(1) Neural Networks
Keyword(2) Bayesian Inference
Keyword(3) Markov Chain Monte Carlo
Keyword(4) sampling efficiency
1st Author's Name Koki Sato
1st Author's Affiliation Hiroshima University(Hiroshima Univ.)
2nd Author's Name Daniel Andrade
2nd Author's Affiliation Hiroshima University(Hiroshima Univ.)
Date 2023-12-21
Paper # IBISML2023-36
Volume (vol) vol.123
Number (no) IBISML-311
Page pp.pp.37-41(IBISML),
#Pages 5
Date of Issue 2023-12-13 (IBISML)