Sugiyama Lab, NII

Computation, Structure, and Learning

Our group, led by Assoc. Prof. Mahito Sugiyama, studies machine learning at National Institute of Informatics (NII) in Tokyo since 2017.

Machine learning is the discipline of studying intelligent computing systems, which nowadays plays a key role in artificial intelligence (AI). To reveal the nature of learning, we focus on the relationship between computational processes, discrete structures, and machine learning models.

We belong to the Department of Informatics, SOKENDAI (The Graduate University for Advanced Studies), which offers 3-year and 5-year Ph.D. courses. More information is available at the NII website.

News

  • Julien Closson from ISIMA joined our lab as an internship student (Apr. 1, 2024).
  • Tatsuki Ebisawa and Kentaro Nakata joined our lab as new Ph.D. students (Apr. 1, 2024).
  • James Enouen from University of Southern California joined our lab as an internship student (Mar. 5, 2024).
  • Chi-Hsien Chang from National Taiwan Univeristy joined our lab as an internship student (Feb. 26, 2024).
  • Our paper "Bringing Structure to Naturalness: On the Naturalness of ASTs" (First author: Profir-Petru Pârțachi) has been accepted to ICSE 2024 Posters Track (Jan. 23, 2024).  
  • Our two presentations (Presenters: Kazu Ghalamkari and Ryuichi Kanoh) received 2023 JSAI Annual Conference Award (Nov. 1, 2023).   [Link (in Japanese)]
  • Zhanpeng Zhou from Shanghai Jiao Tong University joined our lab as an internship student (Sep. 27, 2023).
  • Masatsugu Yamada received his Ph.D. degree in Informatics from SOKENDAI. Congratulations! (Sep. 26, 2023)
  • Our paper "Many-body Approximation for Non-negative Tensors" (First author: Kazu Ghalamkari) has been accepted to NeurIPS 2023 (Sep. 22, 2023).   [arXiv]
  • Our paper "Unsupervised Tensor Based Feature Extraction from Multivariate Time Series" (First author: Kiyotaka Matsue) has been accepted to IEEE Access (Sep. 21, 2023).

▼ more news...

  • Two papers have been accepted to ICML 2023 workshops (Differentiable Almost Everything and Duality Principles for Modern ML) (Jun. 22, 2023).
  • Our paper "How Graph Features from Message Passing Affect Graph Classification and Regression?" (First author: Masatsugu Yamada) has been accepted to Intelligent Data Analysis (Jun. 13, 2023).
  • Our paper "Molecular Graph Generation by Decomposition and Reassembling" (First author: Masatsugu Yamada) has been accepted to ACS Omega (Apr. 10, 2023).   [arXiv]
  • Yuichi Maruyama joined our lab as a new Ph.D. student (Apr. 1, 2023).
  • Our paper "Artificial Neural Network Encoding of Molecular Wavefunctions for Quantum Computing" (First author: Masaya Hagai) has been accepted to Digital Discovery journal (Mar. 30, 2023).   [Paper] [Press Release (in Japanese)]
  • Stefan Mautner from University of Freiburg joined our lab as an internship student (Mar. 27, 2023).
  • Kazu Ghalamkari received his Ph.D. degree in Informatics from SOKENDAI and was awarded Dean's Award! Congratulations! (Mar. 24, 2023)
  • Our paper "Non-negative Low-rank Approximations for Multi-dimensional Arrays on Statistical Manifold" (First author: Kazu Ghalamkari) has been accepted to Information Geometry journal (Jan. 26, 2023).   [Paper]
  • Our paper "Analyzing Tree Architectures in Ensembles via Neural Tangent Kernel" (First author: Ryuichi Kanoh) has been accepted to ICLR 2023 (Jan. 21, 2023).   [arXiv]  [OpenReview]  [Poster]
  • Kazu Ghalamkari has presented his work "Non-negative low-rank approximations for multi-dimensional arrays on statistical manifold" at IG4DS conference (Sep. 23, 2022).  [Slide]
  • Our JST CREST proposal "Machine Learning That Connects to Symbolic Reasoning" (Co-investigators: Katsumi Inoue at NII, Masaaki Nishino at NTT, and Ryosuke Kojima at KyotoU) has been accepted to "Trusted AI" research area (Sep. 20, 2022).
  • Our paper "A Drive-by Bridge Inspection Framework Using Non-Parametric Clusters over Projected Data Manifolds" (First author: Prasad Cheema, Corresponding author: Mehrisadat Makki Alamdari) has been accepted to Mechanical Systems and Signal Processing (MSSP) (Jun. 2, 2022).   [Paper]
  • Prasad Cheema joined our lab as a postdoc researcher (May 1, 2022).
  • Profir-Petru Partachi joined our lab as a postdoc researcher (Apr. 16, 2022).
  • Ryunosuke Ishizaki joined our lab as a new Ph.D. student (Apr. 1, 2022).
  • Kiyotaka Matsue received his Ph.D. degree in Informatics from SOKENDAI. Congratulations! (Mar. 22, 2022).
  • Our paper "A Neural Tangent Kernel Perspective of Infinite Tree Ensembles" (First author: Ryuichi Kanoh) has been accepted to ICLR 2022 (Jan. 21, 2022).   [arXiv]  [OpenReview]  [Poster]
  • Our paper "Fast Rank-1 NMF for Missing Data with KL Divergence" (First author: Kazu Ghalamkari) has been accepted to AISTATS 2022 (Jan. 18, 2022).   [Paper]  [arXiv]  [Code]  [Poster]
  • Our paper "Fast Tucker Rank Reduction for Non-Negative Tensors Using Mean-Field Approximation" (First author: Kazu Ghalamkari) has been accepted to NeurIPS 2021 (Sep. 29, 2021).   [arXiv]  [Paper]  [Code]  [Slide]  [Poster]
  • Our paper "Unsupervised Feature Extraction from Multivariate Time Series for Outlier Detection" (First author: Kiyotaka Matsue) has been accepted to Intelligent Data Analysis (Sep. 11, 2021).
  • Our paper "Investigating Overparameterization for Non-Negative Matrix Factorization in Collaborative Filtering" (First author: Yuhi Kawakami) has been accepted to RecSys 2021 LBR track (July 31, 2021).   [Paper]
  • Our paper "Unsupervised Tensor based Feature Extraction and Outlier Detection for Multivariate Time Series" (First author: Kiyotaka Matsue) has been accepted to DSAA 2021 (July 26, 2021).
  • Our paper "Hierarchical Probabilistic Model for Blind Source Separation via Legendre Transformation" (First author: Simon Luo) has been accepted to UAI 2021 (May 13, 2021).   [arXiv]
  • Our paper "The Volume of Non-Restricted Boltzmann Machines and Their Double Descent Model Complexity" (by P. Cheema & M. Sugiyama) won Best Paper Award at the Deep Learning through Information Geometry workshop at NeurIPS 2020! (Dec. 12, 2020).
  • Seven papers have been accepted to NeurIPS 2020 workshops (4 at DL-IG, 2 at DiffGeo4DL, 1 at BeyondBackProp).
  • Assoc. Prof. Sugiyama will give a talk about "Search for Statistically Significant Interactions" at the 41st IBISML (Oct. 22, 2020).  [Slide]
  • Kazu Ghalamkari, Ryuichi Kanoh, and Yuhi Kawakami joined our lab as new Ph.D. students (Apr. 1, 2020).
  • Ian Andrew Walker joined our lab as an internship student (Jan. 14, 2020).
  • Assoc. Prof. Sugiyama will give a lecture about "Mechanisms of Machine Learning" at National Institute of Technology, Kurume College (Dec. 6, 2019).  [Slide] (in Japanese)
  • Assoc. Prof. Sugiyama will give a talk about "Learning with Incidence Algebra and Dually Flat Structure" at IBIS 2019 (Nov. 20, 2019).  [Slide]
  • Prasad Cheema joined our lab as an internship student (Nov. 18, 2019).
  • Masatsugu Yamada joined our lab as a new Ph.D. student (Oct. 7, 2019).
  • Assoc. Prof. Sugiyama has been awarded "Reviewer Award" of the ECMLPKDD 2019 Journal Track (Sep. 12, 2019).
  • Assoc. Prof. Sugiyama will organize the session "Machine Learning and Discrete Mathematics" at IBIS 2019 (Nov. 20-23, 2019). All slides are available at the IBIS 2019 website.
  • Assoc. Prof. Sugiyama has presented the work "Finding Statistically Significant Interactions between Continuous Features" at IJCAI 2019 (Aug. 10-16, 2019).  [Paper]  [Slide]  [Poster]
  • The lab webpage has been released (Jul. 4, 2019).

Contact

National Institute of Informatics,
2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo 101-8430, Japan
https://www.nii.ac.jp/en/about/access/
Email: mahito (at) nii.ac.jp

© Mahito Sugiyama 2024