Predicting human behavior by combining brain and behavioral measurements

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K20355
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: Institute of Physical and Chemical Research
• Principal Investigator: Morioka Hiroshi 国立研究開発法人理化学研究所, 革新知能統合研究センター, 特別研究員 (20739552)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 機械学習 / 計算神経科学 / 時系列予測 / 非線形解析 / 深層学習

Outline of Final Research Achievements

We developed a novel analysis framework for general nonlinear dynamics. The conventional frameworks generally assumed linear dynamics or additive innovation models, which can be restrictive for their applications to complex dynamics including the human brain. Our new framework, unsupervised representation learning based on deep learning with theoretical justification, is designed for such general dynamical models, and gives very general tools for analyzing them.

Free Research Field

情報学

Academic Significance and Societal Importance of the Research Achievements

一般形の非線形ダイナミクスは不定性が高いことが知られており，同定性を保証した解析法はこれまで存在しなかった．提案法はそのような一般形非線形ダイナミクスの解析を理論的な保証を与えた上で実現するものであり，大きな学術的意義がある．また，提案法は汎用性の高いものであり，様々な対象に広く適用可能なものである．今後様々な理論的拡張が考えられるなど，社会的意義も大きい．