Presentation 2020-01-09
Dimensionality reduction method for gaussian process posteriors based on information geometry
Hideaki Ishibashi, Shotaro Akaho,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper proposes an extension of principal component analysis for gaussian process posteriors which is denoted by GP-PCA. GP-PCA can be applied to multi-task learning, meta-learning and transfer learning. The issue is how to define an structure of a set of GP posteriors such as a coordinate system and a distance. In this study, we define infinite dimensional structure reduced to finite dimensional structure based on information geometry. Especially, we show that a set of GP posteriors becomes a finite dimensional dually flat. Moreover, we demonstrate the effectiveness of GP-PCA through experiments.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) information geometry / gaussian process / multi-task learning / meta-learning / transfer learning
Paper # IBISML2019-20
Date of Issue 2020-01-02 (IBISML)

Conference Information
Committee IBISML
Conference Date 2020/1/9(2days)
Place (in Japanese) (See Japanese page)
Place (in English) ISM
Topics (in Japanese) (See Japanese page)
Topics (in English) Machine learning, etc.
Chair Hisashi Kashima(Kyoto Univ.)
Vice Chair Masashi Sugiyama(Univ. of Tokyo) / Koji Tsuda(Univ. of Tokyo)
Secretary Masashi Sugiyama(Nagoya Inst. of Tech.) / Koji Tsuda(AIST)
Assistant Tomoharu Iwata(NTT) / Shigeyuki Oba(Kyoto Univ.)

Paper Information
Registration To Technical Committee on Infomation-Based Induction Sciences and Machine Learning
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Dimensionality reduction method for gaussian process posteriors based on information geometry
Sub Title (in English)
Keyword(1) information geometry
Keyword(2) gaussian process
Keyword(3) multi-task learning
Keyword(4) meta-learning
Keyword(5) transfer learning
1st Author's Name Hideaki Ishibashi
1st Author's Affiliation Kyushu Institute of Technology(Kyutech)
2nd Author's Name Shotaro Akaho
2nd Author's Affiliation National Institute of Advanced Industrial Science and Technology/RIKEN(AIST/RIKEN)
Date 2020-01-09
Paper # IBISML2019-20
Volume (vol) vol.119
Number (no) IBISML-360
Page pp.pp.17-24(IBISML),
#Pages 8
Date of Issue 2020-01-02 (IBISML)