2023 Fiscal Year Final Research Report
Establishing a practical learning theory based on the analysis of learning from easy data
Project/Area Number |
19H04067
|
Research Category |
Grant-in-Aid for Scientific Research (B)
|
Allocation Type | Single-year Grants |
Section | 一般 |
Review Section |
Basic Section 60010:Theory of informatics-related
|
Research Institution | Kyushu University |
Principal Investigator |
Takimoto Eiji 九州大学, システム情報科学研究院, 教授 (50236395)
|
Co-Investigator(Kenkyū-buntansha) |
畑埜 晃平 九州大学, 基幹教育院, 准教授 (60404026)
|
Project Period (FY) |
2019-04-01 – 2023-03-31
|
Keywords | 計算学習理論 / オンライン意思決定 / 情報圧縮 / バンディット問題 / 決定ダイアグラム / 組合せ最適化 |
Outline of Final Research Achievements |
Introduceing various easiness measures of data, we worked on various topics with the aim of establishing a theory that provides more precise theoretical guarantees for the performance of learning algorithms. We obtained the following main results. (1) Using the data compression rate as an easiness measure, we developed learning algorithms and combinatorial optimization algorithms that operate faster as the compression rate increases. (2) We developed various online decision making algorithms using matrix rank and margin as easiness measures. (3) We showed that any of a wide class of learning problems is reducible to a specific learning problem in the sense that generalization performance (that is, the easiness of data to learn) is preserved.
|
Free Research Field |
計算学習理論
|
Academic Significance and Societal Importance of the Research Achievements |
圧縮データ上の学習や最適化に関する成果は,圧縮が単にメモリの節約になるだけでなく,計算効率も向上することにつながることを示したという点で意義が高い.また,圧縮データに基づき拡張定式化を自動生成する手法は,極めて汎用性が高く,おそらく世界初のものである.オンライン意思決定に関する成果は,商品推薦システムにおいて,商品間の類似関係と顧客の購買傾向に相関がある場合に,高い確度で,顧客が欲する商品を提示することができることを意味しており,データ容易性に自然で有用な解釈を与えたという意味でも意義が高い.学習問題間の還元に関する成果は,学習容易性還元という新たな概念を生み出し,今後の展開が期待できる.
|