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

Basic research on interpretability and causality in modeling time-dependent phenomena

Research Project

Project/Area Number 22K21278
Research Category

Grant-in-Aid for Research Activity Start-up

Allocation TypeMulti-year Fund
Review Section 1001:Information science, computer engineering, and related fields
Research InstitutionKobe University

Principal Investigator

Komatsu Mizuka  神戸大学, システム情報学研究科, 助教 (80856766)

Project Period (FY) 2022-08-31 – 2024-03-31
Project Status Completed (Fiscal Year 2023)
Budget Amount *help
¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
Fiscal Year 2023: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2022: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
KeywordsSciML / 深層学習 / 応用代数 / 微分方程式 / 時系列データ / データ駆動型モデリング / 代数的可観測性 / ニューラルネットワーク / 解釈性 / 状態空間モデル / 代数 / 因果 / 解釈可能性
Outline of Research at the Start

近年,計測技術や機械学習技術の進歩に伴い,データ駆動型のモデリングが多数提案されている.このようなモデリングによる科学的知見の抽出は,個別の問題において試みられてはいるものの,体系的な研究は少ない.そこで,本研究では,現象に関する意味が与えられた解釈可能なモデルパラメータや,現象に内在する因果の扱いに焦点をおき,データからこれらを抽出するためのモデリングに関する,代数に基づく基礎研究を行う.

Outline of Final Research Achievements

In recent years, research utilizing techniques from both machine learning and computational science has been conducted in the interdisciplinary field. However, there is room for advancement particularly in the interpretability of models and treatment of causality in phenomena within this domain. Based on this, in this study, several approaches were developed with a focus on these aspects of modeling aimed at acquiring scientific insights into observed time-series data.
In broad terms, two researches were conducted as follows:
parameter estimation methods based on algebraic techniques for the case where the governing equations of the phenomenon are at least partially known and deep learning-based approaches are effective when the equations governing the phenomena are unknown.

Academic Significance and Societal Importance of the Research Achievements

時系列データから現象に関する知見の獲得を目指す場合,用いられるモデルとして,支配方程式のような解釈性の高いモデルと,深層学習ベースの解釈性の低いモデルがある.前者に関して,モデルの構造によっては,データからパラメータが一意に定まらず,解析結果の信憑性が担保されないという問題がある.本研究では,これを回避すべく代数に基づく推定手法を提案した.後者に関して,時系列グラフ等を導入し解釈性を向上した.

Report

(3 results)
  • 2023 Annual Research Report   Final Research Report ( PDF )
  • 2022 Research-status Report
  • Research Products

    (15 results)

All 2024 2023 2022 Other

All Int'l Joint Research (2 results) Presentation (9 results) (of which Int'l Joint Research: 3 results,  Invited: 4 results) Book (2 results) Remarks (1 results) Funded Workshop (1 results)

  • [Int'l Joint Research] Ecole Polytechnique (LIX)(フランス)

    • Related Report
      2023 Annual Research Report
  • [Int'l Joint Research] Ecole polytechniqu(フランス)

    • Related Report
      2022 Research-status Report
  • [Presentation] 個性の定量化とModel Identifiability2024

    • Author(s)
      小松瑞果
    • Organizer
      第6回日本メディカルAI学会
    • Related Report
      2023 Annual Research Report
    • Invited
  • [Presentation] Application of differential elimination for physics-informed neural networks2024

    • Author(s)
      Mizuka Komatsu
    • Organizer
      XII. Conference on Differential Algebra and Related Topics
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research / Invited
  • [Presentation] Tysonらのモデルに基づく馴化のシミュレーション及び代数的解析2024

    • Author(s)
      中井空
    • Organizer
      日本応用数理学会第9回学生研究発表会
    • Related Report
      2023 Annual Research Report
  • [Presentation] DeepONetを用いた馴化のモデリング2024

    • Author(s)
      安井賢俊
    • Organizer
      日本応用数理学会第9回学生研究発表会
    • Related Report
      2023 Annual Research Report
  • [Presentation] Algebraic design of physical computing system for time-series generation2023

    • Author(s)
      Mizuka Komatsu
    • Organizer
      NeurIPS 2023 Workshop on ML with New Compute Paradigms (MLNCP)
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Algebraic approaches to quantitative modeling of dynamic biological systems2023

    • Author(s)
      Mizuka Komatsu
    • Organizer
      OKO International Symposium 2023
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research / Invited
  • [Presentation] Differential algebraic analysis of state space models and its applications2023

    • Author(s)
      Mizuka Komatsu
    • Organizer
      超幾何学校2023(ワークショップ)
    • Related Report
      2023 Annual Research Report
    • Invited
  • [Presentation] Physics-Informed Neural Networksに対する代数的アプローチおよび疫学モデリングへの応用2023

    • Author(s)
      小松瑞果
    • Organizer
      2023年度数理生物学会年会
    • Related Report
      2023 Annual Research Report
  • [Presentation] 未観測変数をもつPhysics-Informed Neural Networksに関する代数的考察2022

    • Author(s)
      小松瑞果
    • Organizer
      日本応用数理学会 環瀬戸内応用数理研究部会 第26回シンポジウム
    • Related Report
      2022 Research-status Report
  • [Book] 数理科学2024年3月号 <<グレブナー基底>>のすすめ 理論と実践が織り成す数理の世界2024

    • Author(s)
      青木敏,大杉英史, 野呂正行, 土谷昭善,間野修平,小松瑞果,谷口隆晴, 松原宰栄,高山信毅,篠原直行,伊藤琢真,黒川貴司,荒井迅,龍田真
    • Total Pages
      100
    • Publisher
      サイエンス社
    • Related Report
      2023 Annual Research Report
  • [Book] 酵素工学ニュース Vol. 882022

    • Author(s)
      小松瑞果
    • Total Pages
      4
    • Publisher
      酵素工学研究会
    • Related Report
      2022 Research-status Report
  • [Remarks] 共同研究者の個人ウェブページ

    • URL

      http://www.lix.polytechnique.fr/Labo/Gleb.POGUDIN/

    • Related Report
      2022 Research-status Report
  • [Funded Workshop] International Conference on Scientific Computing and Machine Learning 20242024

    • Related Report
      2023 Annual Research Report

URL: 

Published: 2022-09-01   Modified: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi