2021 Fiscal Year Final Research Report
Information geometrical hierarchical modeling
| Project/Area Number |
20K19865
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 61030:Intelligent informatics-related
|
|---|
| Research Institution | Kyushu Institute of Technology |
|---|
Principal Investigator |
Ishibashi Hideaki 九州工業大学, 大学院生命体工学研究科, 助教 (30838389)
|
|---|
| Project Period (FY) |
2020-04-01 – 2022-03-31
|
| Keywords | 情報幾何学 / 階層モデリング / メタ学習 / マルチタスク学習 / ガウス過程 |
|---|
| Outline of Final Research Achievements |
The purpose of this study is to develop a theory of hierarchical modeling for Bayes posteriors based on information geometry. For this purpose, we addressed the following three themes.(1)To define the structure of the set of Bayes posteriors having infinite-dimensional model parameter. (2)Development of a manifold modeling method for a set of Bayes posteriors based on kernel smoother. (3) Development of a hierarchical modeling method for a set of Bayes posteriors with latent variables.
|
| Free Research Field |
機械学習
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究では情報幾何学に基づいて汎用的に利用できるメタ学習，マルチタスク学習，転移学習の方法論を構築した． 特に本研究の枠組みでは一般的に扱われる教師あり学習のメタ学習やマルチタスク学習だけでなく教師なし学習のメタ学習やマルチタスク学習も統一的に扱うことが可能となる． これにより様々なデータの形式や学習タスクに適した学習アルゴリズムをシームレスに提供できるようになった．
|
