Takashi TAKAHASHI / 髙橋昂

○: (たかし)

×: (すばる)

Assistant professor at the Kabashima group in Institute for Physics of Intelligence, The University of Tokyo

takashi-takahashi atmark g.ecc.u-tokyo.ac.jp


I am an assistant professor at the Kabashima group. I work in the fields of statistical mechanics of disordered systems and statistical inference/learning.

See Researchmap for more formal information. (経歴等の正式な情報については Researchmap をご参照ください。)

Papers


scholar

    2024


  1. Koki Okajima, Takashi Takahashi,
    "Asymptotic Dynamics of Alternating Minimization for Non-Convex Optimization"
    arXiv:2402.04751 arXiv
    preprint

  2. 2023


  3. Koki Okajima, Xiangming Meng, Takashi Takahashi, Yoshiyuki Kabashima
    "Average case analysis of Lasso under ultra-sparse conditions"
    Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, PMLR 206:11317-11330, 2023 arXiv
    Peer-reviewed conference paper

  4. Takashi Takahashi
    "Role of Bootstrap Averaging in Generalized Approximate Message Passing"
    2023 IEEE International Symposium on Information Theory (ISIT) (pp.767-772). arXiv
    Peer-reviewed conference paper

  5. 2022


  6. Siqi Na, Tianyao Huang, Yimin Liu, Takashi Takahashi, Yoshiyuki Kabashima, Xiqin Wang
    "Compressed sensing radar detectors under the row-orthogonal design model: a statistical mechanics perspective"
    IEEE Transactions on Signal Processing arXiv
    Refereed journal paper

  7. Takashi Takahashi
    "Sharp Asymptotics of Self-training with Linear Classifier"
    arXiv:2205.07739 arXiv
    preprint

  8. Takashi Takahashi and Yoshiyuki Kabashima
    "Macroscopic Analysis of Vector Approximate Message Passing in a Model-Mismatched Setting"
    IEEE Transactions on Information Theory arXiv
    Refereed journal paper (longer version of the coference paper presented in ISIT2020),

  9. Yuta Nakamura, Takashi Takahashi and Yoshiyuki Kabashima
    "Statistical mechanics analysis of general multi-dimensional knapsack problems",
    J.Phys.A (22, July. 2022.) arXiv
    Refereed journal paper

  10. 2021


  11. Yoshihiko Nishikawa, Jun Takahashi and Takashi Takahashi,
    "Stationary Bootstrap: A Refined Error Estimation for Equilibrium Time Series",
    arXiv:2112.11837 arXiv
    preprint

  12. 2020


  13. Takashi Takahashi and Yoshiyuki Kabashima,
    "Semi-analytic approximate stability selection for correlated data in generalized linear models",
    J. Stat. Mech. (2020) 093402. arXiv
    Refereed journal paper

  14. Takashi Takahashi and Yoshiyuki Kabashima,
    "Macroscopic Analysis of Vector Approximate Message Passing in a Model Mismatch Setting",
    2020 IEEE International Symposium on Information Theory (ISIT) (pp.1403-1408). arXiv
    Peer-reviewed conference paper

  15. 2019


  16. Takashi Takahashi and Yoshiyuki Kabashima,
    "Replicated Vector Approximate Message Passing For Resampling Problem",
    arXiv:1905.09545. arXiv
    Preprint

  17. 2018


  18. Takashi Takahashi and Yoshiyuki Kabashima,
    "A statistical mechanics approach to de-biasing and uncertainty estimation in LASSO for random measurements",
    J. Stat. Mech. (2018) 073405. (open access) arXiv demo
    Refereed journal paper

  19. 2015


  20. Takashi Takahashi and Koji Hukushima,
    "Evidence of a one-step replica symmetry breaking in a three-dimensional Potts glass model",
    Physical Review E 91, 020102(R) (2015). arXiv
    Refereed journal paper

top


Presentations


International Conferences


    2023


  1. (Oral, invited)
    Takashi Takahashi,
    "Exploring bagging with structured data: Insights from precise asymptotics",
    Workshop on Learning and Inference from Structured Data: Universality, Correlations and Beyond | (smr 3850), July 3-7, 2023, ICTP, Trieste, Italy

  2. (Oral, Peer reviewd)
    Takashi Takahashi,
    "Role of Bootstrap Averaging in Generalized Approximate Message Passing",
    2023 IEEE International Symposium on Infomation Theory, June 25-30, 2023, Taipei, Taiwan

  3. 2022


  4. (Oral)
    Takashi Takahashi,
    "Sharp Asymptotics of Self-training with Linear Classifier",
    Youth in High-Dimensions: Recent Progress in Machine Learning, High-Dimensional Statistics and Inference | (smr 3719), June 27-30, 2022, ICTP, Trieste, Italy

  5. 2020


  6. (Oral, Peer reviewed)
    Takashi Takahashi and Yoshiyuki Kabashima,
    "Macroscopic Analysis of Vector Approximate Message Passing in a Model Mismatch Setting",
    2020 IEEE International Symposium on Information Theory, June 21-26, 2020, Los Angeles, California, USA (virtual online conference on the same dates due to the COVID-19 Pandemic)

  7. 2019


  8. (Oral)
    Takashi Takahashi and Yoshiyuki Kabashima,
    "Replicated vector approximate message passing for resampling problem",
    Statistical Physics and Neural Computation (SPNC-2019), Oct 4-6, 2019, Sun Yat-sen University, Guangzhou, China

  9. (Poster)
    Takashi Takahashi and Yoshiyuki Kabashima,
    "Replicated vector approximate message passing for resampling problem",
    40 years of Replica Symmetry Breaking, A conference about systems with many states, 10-13 September 2019, Sapienza University of Rome, Italy

  10. 2014


  11. (Poster)
    Takashi Takahashi and Koji Hukushima,
    "Evidence of one-step replica symmetry breaking in a finite-dimensional Potts glass model",
    Spin glasses: An old tool for new problems, Institute D'Etudes Scientifiques de Cargese, Cargese, France, 2014/8/25-9/6

top

Domestic Conferences


    2023


  1. (Oral)
    Takashi Takahashi
    「高次元モデルにおける不均衡データ分類」,
    (Label-imbalanced classification in a high-dimensional setting),
    16aB202-8, JPS 78th Annual meeting, Tohoku University, Sendai, September. 2023

  2. (Oral)
    Tomohiro Komiyama, Takashi Takahashi and Yoshiyuki Kabashima
    「標本共分散行列の固有空間の解析」,
    (Analyzing eigenspace of sample covariance matrix),
    25aL2-3, JPS Spring meeting, Online, March. 2023

  3. (Oral)
    Takashi Takahashi
    「ブートストラップ平均化された不偏推定量の統計力学的解析」,
    (Bootstrapping debiased lasso: Asymptotic analysis),
    25aL2-4, JPS Spring meeting, Online, March. 2023

  4. (Oral)
    Koki Okajima, Xiangming Meng, Takashi Takahashi and Yoshiyuki Kabashima
    「極スパース条件におけるLasso によるサポート復元条件」,
    (On the conditions for support recovery using Lasso under ultra sparse settings),
    22pL2-10, JPS Spring meeting, Online, March. 2023

  5. (Oral)
    Yuta Nakamura, Takashi Takahashi and Yoshiyuki Kabashima
    「一般化されたナップザック問題の統計力学的解析」,
    (Statistical mechanics analysis of generalized knapsack problems),
    22pL2-6, JPS Spring meeting, Online, March. 2023

  6. 2022


  7. (Poster)
    髙橋 昂,
    「反復型自己学習アルゴリズムの典型性能評価」,
    第25回情報論的学習理論ワークショップ (IBIS 2022),, 21-22, November 2022, Tsukuba, Japan

  8. (Oral)
    Koki Okajima, Xiangming Meng, Takashi Takahashi and Yoshiyuki Kabashima
    「極スパース状況下におけるLASSO 回帰の統計力学的解析」,
    (Statistical mechanical analysis of LASSO regression under ultra sparse settings),
    12pH112-3, JPS fall meeting, Tokyo, September. 2022

  9. (Oral)
    Takashi Takahashi
    「反復型自己学習アルゴリズムのレプリカ解析」,
    (A statistical mechanics analysis of iterative self-training),
    14pH112-12, JPS fall meeting, Tokyo, September. 2022

  10. (Oral)
    Satoshi Ishii, Takashi Takahashi, Tomoyuki Obuchi, Ayaka Sakata and Yoshiyuki Kabashima
    「線形制約下におけるSCAD最小化問題の解について」,
    (On solutions of SCAD minimization problem under linear constraint),
    15aB14-2, 77th JPS meeting, Online, March. 2022

  11. (Oral)
    Takashi Takahashi
    「高次元統計領域における対角スケーリングの統計力学的解析」
    (A statistical mechanics analysis of diagonal scaling under high-dimensional limit)
    15aB14-3, 77th JPS meeting, Online, March. 2022

  12. (Oral)
    Yusuke Morita, Takashi Takahashi and Yoshiyuki Kabashima
    「標本分散共分散行列の最大固有値に関する統計力学的解析」,
    (statistical mechanics analysis of largest eigenvalue of sample variance-covariance matrix),
    15aB14-8, 77th JPS meeting, Online, March. 2022

  13. 2021


  14. (Oral)
    Yuta Nakamura, Takashi Takahashi and Yoshiyuki Kabashima
    「個数制限のないナップザック問題の統計力学的解析」,
    (Statistical mechanics analysis of unbounded knapsack problems)
    23pL4-9, JPS Fall meeting, Online, Sep. 2021.

  15. (Oral)
    Satoshi Ishii, Takashi Takahashi, Tomoyuki Obuchi, Ayaka Sakata and Yoshiyuki Kabashima
    「線形制約下における SCAD 最小化問題の解について」,
    (On solutions of SCAD minimization problem under linear constraint)
    23pL4-12, JPS Fall meeting, Online, Sep. 2021.

  16. (Oral)
    Yusuke Morita, Takashi Takahashi and Yoshiyuki Kabashima
    「標本分散共分散行列の漸近固有値分布に関する統計力学的解析」,
    (Statistical mechanics analysis of asymptotic eigenvalue distribution of sample covariance matrices)
    23pL4-14, JPS Fall meeting, Online, Sep. 2021.

  17. (Oral)
    Takashi Takahashi
    「半教師あり学習の平衡統計力学的解析」,
    (An equilibrium statistical mechanical analysis of semi-supervised learning)
    23pL4-13, JPS Fall meeting, Online, Sep. 2021.

  18. (Oral)
    Takashi Takahashi
    「安定性選択法の典型性能評価」,
    (Typical performance analysis of stability selection)
    13pL2-9 , 76th JPS meeting, Online, Mar. 2021.

  19. (Poster)
    Yoshihiko Nishikawa, Jun Takahashi, and Takashi Takahashi
    「ブートストラップ法による定常時系列データの信頼性評価」,
    (Bootstrapping stationary time series data for reliability analysis)
    PSL-30, 76th JPS meeting, Online, Mar. 2021.

  20. 2020


  21. (Oral, invited)
    Takashi Takahashi
    「高次元統計学におけるリサンプリング法に対する統計力学的アプローチ」,
    Random Matrices, Free Probability, and Machine Learning(ランダム行列と自由確率と機械学習),

  22. (Oral)
    Takashi Takahashi and Yoshiyuki Kabashima,
    「モデル不整合時の一般化線形モデルの解析とリサンプリングへの応用」,
    (A statistical mechanical analysis of generalized linear models in a model mismatch setting and its application to resampling methods)
    9aL2-4, JPS Fall meeting, Online, Sep. 2020.

  23. (Oral)
    Takashi Takahashi and Yoshiyuki Kabashima,
    「モデル不整合時におけるベクトル近似メッセージパッシングアルゴリズムの解析」,
    (An analysis of vector approximate message passing in a model mismatched setting)
    17aK43-10, JPS meeting, Online, Mar. 2020,

  24. 2019


  25. (Poster)
    髙橋 昂, 樺島 祥介,
    「相関のある特徴に対する半解析的stability selection法の開発」,
    第22回情報論的学習理論ワークショップ (IBIS 2019), 20-21, November 2019, Nagoya, Japan

  26. (Oral)
    髙橋 昂, 樺島 祥介,
    「相関のある特徴量に対する半解析的Stability Selection法」,
    科研費シンポジウム 「統計学と機械学習の数理と展開」, 18-19, September 2019, Tokyo Institute of Technology, Japan

  27. (Oral)
    Takashi Takahashi and Yoshiyuki Kabashima,
    「相関のある特徴パターンに対する半解析的stability selection法」,
    (Semi-analytic stability selection for correlated feature patterns)
    14pG214-4, 74th JPS meeting, Ito, Mar. 2019.

  28. 2018


  29. (Oral)
    Takashi Takahashi and Yoshiyuki Kabashima,
    「適応Thouless-Anderson-Palmer法によるLASSOに関する半解析的パラメトリックブートストラップ法の開発」,
    (Adaptive Thouless-Anderson-Palmer approach to semi-analytic parametric bootstrap method in LASSO)
    9pM203-4, JPS Fall meeting, Kyotanabe, Sep. 2018.,

  30. (Oral)
    Takashi Takahashi and Yoshiyuki Kabashima,
    「適応Tholess-Anderson-Palmer法によるLASSO解の統計的不確かさ評価手法の開発」,
    (Adaptive Thouless-Anderson-Palmer approach to quantification of uncertainty associated with LASSO)
    22aK703-3, 73th JPS meeting, Saitama, Mar. 2018.,

  31. 2015


  32. (Oral)
    Takashi Takahashi and Koji Hukushima,
    「3次元ポッツグラスにおける静的・動的協調スケールの研究」,
    (Numerical study of static and dynamic cooperative scales in a three dimensional Potts glass model)
    24pBL-4, 70th JPS meeting, Tokyo, Mar. 2015.,

  33. (Poster)
    Takashi Takahashi and Koji Hukushima,
    「3次元ポッツグラス模型における1段階レプリカ対称性の破れと動的異常」,
    第4回ソフトマター研究会,P-24, Aichi,Jan. 2015.,

  34. 2014


  35. (Poster)
    Takashi Takahashi and Koji Hukushima,
    「3次元スピングラス模型の結合レプリカ系の性質」,
    (Monte-Carlo study of a coupled replica system of a three dimensional Potts glass model)
    8pPSA-3, JPS Fall meeting, Aichi, Sep. 2014.,

  36. (Oral)
    Takashi Takahashi and Koji Hukushima,
    「有限次元スピングラス模型の示す1段階レプリカ対称性の破れ」,
    (One-step replica symmetry breaking in a finite dimensional spin-glass model)
    28aAR-6, 69th JPS meeting, Kanagawa, Mar. 2014.,

  37. 2013


  38. (Oral)
    Takashi Takahashi and Koji Hukushima,
    「3次元ポッツグラス模型のスピングラス相の性質」,
    (Equilibrium properties of spin-glass phase of 3d potts glass model)
    28pKL-6, JPS Fall meeting, Tokushima, Sep. 2013.,

top

Seminars


    2024


  1. 「疑似ラベルを用いた自己学習の平均場解析」,
    滋賀大学データサイエンスセミナー,
    2024年1月22日(月)10:30-, 滋賀大・講堂多目的室1

  2. 2023


  3. 「疑似ラベルを用いた学習のレプリカ解析」,
    第57回統計的機械学習セミナー / The 57th Statistical Machine Learning Seminar,
    2023年8月28日(月)15:00-, 東大・駒場ファカルティハウス・セミナー室

  4. 「レプリカ法による線形モデルの精密評価」,
    応用統計ワークショップ @ 日本経済国際共同研究センター ,
    2023年4月9日(金)16:50-18:35, 東京大学大学院経済学研究科 学術交流棟 (小島ホール)1階 第1セミナー室

  5. 2022


  6. 「疑似ラベルに基づく半教師あり学習の統計力学的解析」,
    情報数物研究会 @ 東北大学 大学院情報科学研究科 応用情報科学専攻 応用情報技術論講座 物理フラクチュオマティクス論分野 田中・大関研 ,
    2022年1月18日(火)16:00-17:30, Online

  7. 2021


  8. 「安定性選択法の近似計算と典型性能評価」,
    統計物理と統計科学のセミナー @ 統計数理研究所,
    2021年11月16日 (火) 17:00-18:00, Online

  9. 「ブートストラップ法の計算・理論に対する統計物理学的アプローチ」,
    東京大学数理情報学談話会 @ 東京大学大学院情報理工学系研究科数理情報学専攻 ,
    2021年10月25日(月)17:00-18:00, Online

top