Machine Learning Meets Statistical Physics II

Presentations

• Variational gradient descent: enhancing generalization with automatically learned landscape-dependent noise.

  ORAL · Invited

  March 19, 2025, 8:00 AM – March 19, 2025, 8:36 AM

  Presenters

  • David Hathcock

    • IBM Thomas J. Watson Research Center

  Authors

  • David Hathcock

    • IBM Thomas J. Watson Research Center

  • Yuhai Tu

    • IBM Thomas J. Watson Research Center

  View abstract →

• Top-Down approach to dynamical coarse-graining using Differentiable Generalized Langevin Equation

  ORAL

  March 19, 2025, 8:36 AM – March 19, 2025, 8:48 AM

  Publication: Jeong, Jinu, Ishan Nadkarni, and Narayana Aluru. "DiffGLE: Differentiable Coarse-Grained Dynamics using Generalized Langevin Equation." arXiv preprint arXiv:2410.08424 (2024).

  Presenters

  • Ishan Mangesh Nadkarni

    • The University of Texas at Austin

  Authors

  • Ishan Mangesh Nadkarni

    • The University of Texas at Austin

  • Jinu Jeong

    • University of Illinois at Urbana−Champaign, Urbana
    • The University of Illinois at Urbana-Champaign

  • Narayana R Aluru

    • The University of Texas at Austin
    • University of Texas at Austin

  View abstract →

• The Manifold Packing Loss Function: A Physics-Inspired Approach to Contrastive Self-Supervised Learning

  ORAL

  March 19, 2025, 8:48 AM – March 19, 2025, 9:00 AM

  Presenters

  • Guanming Zhang

    • New York University (NYU)

  Authors

  • Guanming Zhang

    • New York University (NYU)

  • David J Heeger

    • New York University

  • Stefano Martiniani

    • New York University (NYU)

  View abstract →

• Mutual Information Can Be Estimated when Undersampled Data Have Low-Dimensional Latent Structure

  ORAL

  March 19, 2025, 9:00 AM – March 19, 2025, 9:12 AM

  Presenters

  • Eslam Abdelaleem

    • Georgia Institute of Technology

  Authors

  • Eslam Abdelaleem

    • Georgia Institute of Technology

  • K. Michael Martini

    • Emory University

  • Ilya M Nemenman

    • Emory University

  View abstract →

• Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks

  ORAL

  March 19, 2025, 9:12 AM – March 19, 2025, 9:24 AM

  Publication: Tianyu He, Darshil Doshi, Aritra Das, Andrey Gromov; "Learning to grok: Emergence of in-context learning and skill compostion in modular arithmetic tasks"; NeurIPS 2024 (Oral)

  Presenters

  • Darshil H Doshi

    • University of Maryland College Park

  Authors

  • Darshil H Doshi

    • University of Maryland College Park

  • Tianyu He

    • University of Maryland College Park

  • Aritra Das

    • University of Maryland, College Park
    • University of Maryland College Park

  • Andrey Gromov

    • University of Maryland College Park

  View abstract →

• Specialization-generalization transition in exemplar-based in-context learning

  ORAL

  March 19, 2025, 9:24 AM – March 19, 2025, 9:36 AM

  Presenters

  • Chase Waring Goddard

    • Princeton University

  Authors

  • Chase Waring Goddard

    • Princeton University

  • Lindsay Maleckar Smith

    • Princeton University

  • Vudtiwat Ngampruetikorn

    • University of Sydney

  • David J Schwab

    • CUNY Graduate Center
    • The Graduate Center, CUNY
    • CUNY

  View abstract →

• Diffusion Models as an Extension of Variational Autoencoders

  ORAL

  March 19, 2025, 9:36 AM – March 19, 2025, 9:48 AM

  Presenters

  • Kentaro Kaba

    • Institute of Science Tokyo

  Authors

  • Kentaro Kaba

    • Institute of Science Tokyo

  • Reo Shimizu

    • Tohoku University

  • Masayuki Ohzeki

    • Graduate School of Information Sciences, Tohoku University, Department of Physics, Institute of Science Tokyo, Sigma-i Co., Ltd.
    • Institute of Science Tokyo, Tohoku University, Sigma-i Co., Ltd.,
    • Graduate School of Information Sciences, Tohoku University; Department of Physics, Institute of Science Tokyo; Sigma-i Co., Ltd.

  • Yuki Sughiyama

    • Tohoku University

  View abstract →

• Origins and mitigation of double descent in sparse sensing

  ORAL

  March 19, 2025, 9:48 AM – March 19, 2025, 10:00 AM

  Publication: Andrei A. Klishin, Samuel E. Otto, J. Nathan Kutz, Krithika Manohar, in preparation (2024)

  Presenters

  • Andrei A. Klishin

    • University of Hawaiʻi at Mānoa

  Authors

  • Andrei A. Klishin

    • University of Hawaiʻi at Mānoa

  • Samuel E Otto

    • Cornell University

  • J. Nathan Kutz

    • University of Washington, AI Institute for Dynamic Systems

  • Krithika Manohar

    • University of Washington

  View abstract →

• Statistical Mechanics of Double Descent in Deep Learning: a Phase Transition Perspective

  ORAL

  March 19, 2025, 10:00 AM – March 19, 2025, 10:12 AM

  Presenters

  • Chan Li

    • University of California, San Diego

  Authors

  • Chan Li

    • University of California, San Diego

  • Nigel Goldenfeld

    • University of California, San Diego

  View abstract →

• LLMs Learn Physical Rules of Dynamical Systems: A Geometric Investigation of Emergent Algorithms

  ORAL

  March 19, 2025, 10:12 AM – March 19, 2025, 10:24 AM

  Publication: T. J.B. Liu, N. Boullé, R. Sarfati, & C. J. Earls, LLMs learn governing principles of dynamical systems, revealing an in-context neural scaling law, EMNLP (2024)
  
  Liu, T.J., Boull'e, N., Sarfati, R., & Earls, C.J. Density estimation with LLMs: a geometric investigation of in-context learning trajectories, (2024)

  Presenters

  • Toni Jianbang Liu

    • Cornell University

  Authors

  • Toni Jianbang Liu

    • Cornell University

  • Raphael Sarfati

    • Cornell University

  • Christopher Earls

    • Cornell University
    • Cornell university

  • Nicolas Boulle

    • Imperial College London

  View abstract →

• Physics-Inspired Model Compression of Neural Networks

  ORAL

  March 19, 2025, 10:24 AM – March 19, 2025, 10:36 AM

  Presenters

  • Daniel T Bernstein

    • Princeton University

  Authors

  • Daniel T Bernstein

    • Princeton University

  • David J Schwab

    • CUNY Graduate Center
    • The Graduate Center, CUNY
    • CUNY

  View abstract →

• Long-range order in classification tasks

  ORAL

  March 19, 2025, 10:36 AM – March 19, 2025, 10:48 AM

  Publication: Zhang, YH., Sipling, C., Qiu, E. et al. Collective dynamics and long-range order in thermal neuristor networks. Nat Commun 15, 6986 (2024). https://doi.org/10.1038/s41467-024-51254-4
  Computing with long-range order: when, why, and how. In preparation.

  Presenters

  • Yuan-Hang Zhang

    • University of California, San Diego

  Authors

  • Yuan-Hang Zhang

    • University of California, San Diego

  • Chesson Sipling

    • University of California, San Diego

  • Massimiliano Di Ventra

    • University of California, San Diego

  View abstract →