Updated on 2026/08/19

 
IMAGAWA Takahisa
 
Scopus Paper Info
Total Paper Count: 10 Total Citation Count: 117 h-index: 5

The bar and line graphs display up to the most recent 20 years.

Affiliation
Faculty of Computer Science and Systems Engineering Department of Intelligent and Control Systems
Job
Assistant Professor
External link

Research Areas

  • Informatics / Intelligent informatics  / Artificial Intelligence, Reinforcement Learning, Planning

Degree

  • 東京大学  -  博士（学術）   2018.03

  • 東京大学  -  修士（学術）   2015.03

  • 東京大学  -  学士（教養）   2013.03

Biography in Kyutech

  • 2024.02
     

    Kyushu Institute of Technology   Faculty of Computer Science and Systems Engineering   Department of Intelligent and Control Systems   Assistant Professor  

Papers

  • Cost-Aware Embedding Dimension Selection via Accuracy and Training-Cost Surrogate Models Reviewed

    Imagawa, Takahisa and Enokida, Shuichi

    18th International Joint Conference IJCCI 2026   2026.10

    　More details

    Authorship：Lead author,　Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

  • 物体検出器の知識蒸留におけるアノテーションコストを考慮した蒸留手法に関する研究

    横石和真, 今川孝久, 榎田修一

    SSII2026   2026.06

    　More details

    Authorship：Corresponding author   Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

  • Exploring Pre-Service Teachers’ Reflection on Nonverbal Behavior in Microteaching Through Three-Point Comparison Feedback Reviewed

    Shirasaka S., Imagawa T., Enokida S.

    Education Sciences   16 ( 5 )   2026.05

    　More details

    Language：English   Publishing type：Research paper (scientific journal)

    DOI： 10.3390/educsci16050760

    Scopus

    Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105040046374&origin=inward

  • MOTIP における入力時系列長に対する ID マッチング精度の予測と その活用による再学習コスト削減手法

    川上舞, 今川孝久, 榎田修一

    CVIM2026年1月研究会   2026.01

    　More details

    Authorship：Corresponding author   Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

  • The double disconnect: why automated facial expression analysis does not predict peer evaluation in microteaching Reviewed

    Shirasaka S., Imagawa T., Enokida S.

    Frontiers in Education   11   2026.01

    　More details

    Language：English   Publishing type：Research paper (scientific journal)

    DOI： 10.3389/feduc.2026.1835461

    Scopus

    Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105046123444&origin=inward

  • Sample-Efficient Reinforcement Learning through Cross Bisimulation-Based Implicit Imitation Learning Reviewed

    IMAGAWA Takahisa, ENOKIDA Shuichi

    Proceedings of the Annual Conference of JSAI ( The Japanese Society for Artificial Intelligence )   JSAI2026 ( 0 )   5M1GS2b02 - 5M1GS2b02   2026.01

    　More details

    Authorship：Lead author   Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

    <p>Reinforcement learning is a useful methodology with a wide range of application. examples; however, it generally requires a large amount of data, and reducing this requirement remains an important challenge.In this study, we propose Cross Bisimulation-based Implicit Imitation Learning (CBI2L), a method that improves the sample efficiency of reinforcement learning by leveraging a small amount of imitation data that is not necessarily high quality.Specifically, for the Markov decision processes of two agents—namely, a mentor to be imitated and an observer that performs reinforcement learning—we define a pseudometric called the cross bisimulation metric, which is based on the difference in cumulative rewards between them.We then theoretically analyze (i) the unique existence of the cross bisimulation metric via the fixed-point theorem and (ii) the relationship between the difference in the expected cumulative rewards of the mentor and observer and the cross bisimulation metric, thereby establishing the validity of CBI2L, which uses the mentor’s rewards for the observer’s learning.Furthermore, we incorporate CBI2L into Soft Actor-Critic (SAC) and empirically show that it improves learning efficiency compared to SAC in the PointMaze environment.</p>

    DOI： 10.11517/pjsai.jsai2026.0_5m1gs2b02

    CiNii Research

  • Biased Exploration Q-Learning: A Simple Method for Embedding Knowledge into Reinforcement Learning Reviewed

    Imagawa T., Enokida S.

    International Conference on Agents and Artificial Intelligence   2   1242 - 1253   2026.01

    　More details

    Authorship：Lead author   Language：English   Publishing type：Research paper (international conference proceedings)

    DOI： 10.5220/0014246600004052

    Scopus

    Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105041701330&origin=inward

  • Implicit Imitation Learningの応用による強化学習の効率化

    今川孝久, 榎田修一

    IBIS2025   2025.11

    　More details

    Authorship：Lead author,　Corresponding author   Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

  • Optimization of Masking Selection Ratio in Diffusion Action Segmentation Reviewed

    Ooyama, Yuta and Imagawa, Takahisa and Enokida, Shuichi

    Proceedings of 8th International Symposium on Future Active Safety Technology towards Zero-Traffic Accidents   2025.09

    　More details

    Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

  • Pose Estimation Using Mask Processing for Robustness Against Occlusions in Overlapping People Reviewed

    Ide, Yuki and Imagawa, Takahisa and Enokida, Shuichi

    Proceedings of 8th International Symposium on Future Active Safety Technology towards Zero-Traffic Accidents   2025.09

    　More details

    Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

  • Efficient Embedding Dimension Number Selection Method for Deep Learning Focusing on Loss Variation in the Early Learning Phase Reviewed

    Uchida, Aoto and Imagawa, Takahisa and Enokida Shuichi

    Proceedings of 8th International Symposium on Future Active Safety Technology towards Zero-Traffic Accidents   2025.09

    　More details

    Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

  • 境界重視型損失関数を追加したHQ-SAM の提案とダクトホース内部検査への応用

    久保田愛梨, 相馬康宏, 相馬貴之, 今川孝久, 榎田修一

    FIT2025   2025.09

    　More details

    Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

  • GroupPose を用いた複数人同時姿勢推定におけるアテンションのマスキングに関する研究

    笹岡亮太, 今川孝久, 榎田修一

    MIRU2025 Extended Abstract集   2025.08

    　More details

    Authorship：Corresponding author   Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

  • ハンドクラフト特徴量による深層学習ベースの三次元物体検出器の性能向上

    平川絢士, 今川孝久, 榎田修一

    DIA2025講演論文集   2025.03

    　More details

    Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

    Kyutacar

  • 事前知識を用いたQ学習

    今川孝久，榎田修一

    IBISML研究会論文集   2024.12

    　More details

    Authorship：Lead author,　Corresponding author   Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

    Kyutacar

  • A Method for Efficiently Selecting the Number of Dimension in Deep LearningFocusing on Change in Loss in the Early Phase of Learning "jointly worked"

    Uchida, Aoto/ Imagawa, Takahisa/ Enokida, Shuichi

    2024.09

    　More details

    Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

    Kyutacar

  • 人物の重なりに対する頑健性向上のためのマスク処理を用いた姿勢推定

    井手宥希，今川孝久，榎田修一

    MIRU2024 Extended Abstract集   2024.08

    　More details

    Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

    Kyutacar

  • Diffusion Action Segmentationにおけるフレーム特徴量へのマスキングに関する研究

    大山佑太，今川孝久，榎田修一

    MIRU2024 Extended Abstract集   2024.08

    　More details

    Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)

    Kyutacar

  • DROPOUT Q-FUNCTIONS FOR DOUBLY EFFICIENT REINFORCEMENT LEARNING Reviewed International journal

    Hiraoka T., Imagawa T., Hashimoto T., Onishi T., Tsuruoka Y.

    ICLR 2022 - 10th International Conference on Learning Representations   2022.01

    　More details

    Authorship：Last author   Language：English   Publishing type：Research paper (international conference proceedings)

    Scopus

    Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146871335&origin=inward

  • Off-Policy Meta-Reinforcement Learning with Belief-Based Task Inference Reviewed International journal

    Imagawa T., Hiraoka T., Tsuruoka Y.

    IEEE Access   10   49494 - 49507   2022.01

    　More details

    Authorship：Lead author   Language：English   Publishing type：Research paper (scientific journal)

    DOI： 10.1109/ACCESS.2022.3170582

    Scopus

    Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85129665822&origin=inward

  • Meta-Model-Based Meta-Policy Optimization Reviewed International journal

    Hiraoka T., Imagawa T., Tangkaratt V., Osa T., Onishi T., Tsuruoka Y.

    Proceedings of Machine Learning Research   157   129 - 144   2021.01

    　More details

    Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

    Scopus

    Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85137101456&origin=inward

  • Learning robust options by conditional value at risk optimization Reviewed International journal

    Hiraoka T., Imagawa T., Mori T., Onishi T., Tsuruoka Y.

    Advances in Neural Information Processing Systems   32   2019.01

    　More details

    Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

    Scopus

    Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85090171999&origin=inward

▼display all

Conference Prsentations (Oral, Poster)

  • Q-Learning with Prior Knowledge

    Takahisa Imagawa

    2024.12 

    　More details

    Event date： 2025.12.20 - 2025.12.21   Language：Japanese   Country：Japan  

Grants-in-Aid for Scientific Research

  • 他者データを学習のバイアスとして活用することによる強化学習の効率化

    Grant number：25K21153   2025.04   若手研究

    　More details

    強化学習は学習者が試行錯誤を繰り返しその結果の良し悪しをもとに学習する方法で，広い応用先を持つ．
    しかし，多くの場合，無数の試行錯誤が必要となり，このことが実応用を妨げる一因となっている．
    本研究では，強化学習をする学習者の他に，他者が存在する状況を想定する．
    その場合，他者が学習者にとって有益な行動をとれば，学習者の学習が大きく進展する可能性がある．
    本研究では他者からの学習を実現するための基礎となる手法を考案する．