• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to project page

2021 Fiscal Year Final Research Report

Theory and Application of Statistical Reinforcement Learning

Research Project

  • PDF
Project/Area Number 17H00757
Research Category

Grant-in-Aid for Scientific Research (A)

Allocation TypeSingle-year Grants
Section一般
Research Field Intelligent informatics
Research InstitutionThe University of Tokyo

Principal Investigator

Sugiyama Masashi  東京大学, 大学院新領域創成科学研究科, 教授 (90334515)

Project Period (FY) 2017-04-01 – 2022-03-31
Keywords強化学習 / 機械学習 / 多腕バンディット問題 / 模倣学習 / ベイズ推論 / ロバスト性
Outline of Final Research Achievements

In this research, we developed theories and algorithms for sqeuential decision making and probabilistic inference. In the study of reinforcement learning, we developed methods for weakly supervised imitation learning and hierarchization of complex problems to improve their practicality, and demonstrated their effectiveness experimentally. For multi-arm bandit problems, we developed algorithms with theoretical guarantees for linear bandit, dueling bandit, good-arm identification, and combinatorial bandit. In the area of probabilistic inference, we have conducted research on making Bayesian inference robust, speeding up approximate computation, and modeling temporal events, and have verified the effectiveness of these methods both theoretically and experimentally.

Free Research Field

知能情報学

Academic Significance and Societal Importance of the Research Achievements

逐次的意思決定や確率的推論は,今後の発展が大いに期待される重要な機械学習技術である.本研究では,強化学習や多腕バンディットの適用範囲を拡大する新しいアルゴリズムを開発するとともに,確率的推論のロバスト性向上や近似計算の高速化に関する研究を行った.このような基礎理論的な研究成果は,逐次的意思決定や確率的推論の原理の解明に貢献するものであり,機械学習分野の主要国際会議で学術的に高い評価を受けた.また,開発したアルゴリズムの有効性は計算機実験によって示されており,将来の社会実装につながる社会的意義のある開発であるとも考えられる.

URL: 

Published: 2023-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi