一覧に戻る

タイトル
  • en Principal Component Analysis for Gaussian Process Posteriors
作成者
    • en Ishibashi, Hideaki ja 石橋, 英朗 ja-Kana イシバシ, ヒデアキ
    • e-Rad 30838389
    • en Akaho, Shotaro
権利情報
  • This is the author’s final version, and that the article has been accepted for publication in Neural Computation.
主題
  • Other Gaussian process
  • Other Information geometry
  • Other Multi-task learning
  • Other Metalearning
  • Other Functional data analysis
内容注記
  • Abstract en This letter proposes an extension of principal component analysis for gaussian process (GP) posteriors, denoted by GP-PCA. Since GP-PCA estimates a low-dimensional space of GP posteriors, it can be used for metalearning, a framework for improving the performance of target tasks by estimating a structure of a set of tasks. The issue is how to define a structure of a set of GPs with an infinite-dimensional parameter, such as coordinate system and a divergence. In this study, we reduce the infiniteness of GP to the finite-dimensional case under the information geometrical framework by considering a space of GP posteriors that have the same prior. In addition, we propose an approximation method of GP-PCA based on variational inference and demonstrate the effectiveness of GP-PCA as meta-learning through experiments.
出版者 Massachusetts Institute of Technology Press
日付
    Issued2022-04-15
言語
  • eng
資源タイプ journal article
出版タイプ AM
資源識別子 HDL http://hdl.handle.net/10228/0002000098 , URI https://kyutech.repo.nii.ac.jp/records/2000098
関連
  • isVersionOf DOI https://doi.org/10.1162/neco_a_01489
収録誌情報
    • PISSN 0899-7667
    • EISSN 1530-888X
      • en Neural Computation
      • 巻34 号5 開始ページ1189 終了ページ1219
ファイル
コンテンツ更新日時 2025-07-14