publications

My Google Scholar profile is here.


Book Chapters

  1. Rethinking importance weighting for transfer learning
    Nan Lu*, Tianyi Zhang, Tongtong Fang, Takeshi Teshima, and Masashi Sugiyama
    In: Federated and Transfer Learning, 2022
    Book
  2. Forecasting internally displaced people’s movements with artificial intelligence
    Ayaka Oishi*, Takeshi Teshima, Kunikazu Akao, Tsuyoshi Kano, Megumi Kiriha, Naoki Kojima, Takuya Sasaki, Kentaro Takahira, Jumpei Takami, Tomonari Takeuchi, Kenji Tajima, Chihiro Noda, Hikaru Hirose, and Shigeki Yamanaka
    In: Digital Innovations, Business and Society in Africa: New Frontiers and a Shared Strategic Vision, 2022
    Book

Journal Articles (Refereed)

  1. Universal approximation property of invertible neural networks
    Isao Ishikawa*, Takeshi Teshima, Koichi Tojo, Kenta Oono, Masahiro Ikeda, and Masashi Sugiyama
    Journal of Machine Learning Research, 2023
    (First two authors contributed equally.)

Conference Proceedings (Refereed)

  1. Incorporating causal graphical prior knowledge into predictive modeling via simple data augmentation
    Takeshi Teshima*, and Masashi Sugiyama
    In: Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence (UAI 2021), PMLR 161, 2021
    (Acceptance rate: 26.4%)
    Code Poster Slides Venue Video
  2. Non-negative Bregman divergence minimization for deep direct density ratio estimation
    Masahiro Kato*, and Takeshi Teshima
    In: Proceedings of the 38th International Conference on Machine Learning (ICML 2021), 2021
    (Acceptance rate: 21.4%)
    Press release
  3. γ-ABC: Outlier-robust approximate bayesian computation based on a robust divergence estimator
    Masahiro Fujisawa*, Takeshi Teshima, Issei Sato, and Masashi Sugiyama
    In: Proceedings of The 24th International Conference on Artificial Intelligence and Statistics (AISTATS 2021), PMLR 130, 2021
    (Acceptance rate: 29.8%)
    Venue
  4. Few-shot domain adaptation by causal mechanism transfer
    Takeshi Teshima*, Issei Sato, and Masashi Sugiyama
    In: Proceedings of the 37th International Conference on Machine Learning (ICML2020), PMLR 119, 2020
    (Acceptance rate: 21.8%)
    Code Slides Session Video
  5. Coupling-based invertible neural networks are universal diffeomorphism approximators
    Takeshi Teshima*, Isao Ishikawa, Koichi Tojo, Kenta Oono, Masahiro Ikeda, and Masashi Sugiyama
    In: Advances in Neural Information Processing Systems 33, 2020
    (First two authors contributed equally. Selected for oral presentation; 105 orals out of 1900 accepted papers. Acceptance rate 20.1%, oral acceptance rate 1.1%)
    Code Slides
  6. Clipped matrix completion: a remedy for ceiling effects
    Takeshi Teshima*, Miao Xu, Issei Sato, and Masashi Sugiyama
    In: Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 2019
    (Acceptance rate: 16.2%)
    Supplementary Code Slides
  7. Learning from positive and unlabeled data with a selection bias
    Masahiro Kato*, Takeshi Teshima, and Junya Honda
    In: Proceedings of the Seventh International Conference on Learning Representations (ICLR 2019), 2018
    (No page numbers were assigned in these proceedings.)

Workshop Papers

  1. Universal approximation property of neural ordinary differential equations
    Takeshi Teshima*, Koichi Tojo, Masahiro Ikeda, Isao Ishikawa, and Kenta Oono
    2020
    (Presented at NeurIPS2020 Workshop: Differential Geometry meets Deep Learning, DiffGeo4DL.)
    Slides Venue Video

Patents and Patent Applications

  1. Sampling unique molecular structures from autoencoders
    Takeshi Teshima, and Hiroshi Kajino
    2020
    (US Patent App. 17/060,862; Publication No. US20220108765A1. Patent pending.)

Theses

  1. Causal Machine Learning from Small Data: A Data Augmentation Approach
    Takeshi Teshima
    The University of Tokyo, 2022
    (UTokyo repository forthcoming)
  2. Research on Restoring Clipped Low-rank Matrices
    Takeshi Teshima
    The University of Tokyo, 2019