APS Logo

Neural Systems III

FOCUS · K02 · ID: 46189

Presentations

  • Associative Memory of Knowledge Structures and Sequences

    ORAL · Invited

    –

    Publication: Julia Steinberg and Haim Sompolinsky. Associative memory of structured knowledge. In preparation

    Presenters

    • Julia A Steinberg

      Princeton University

    Authors

    • Julia A Steinberg

      Princeton University

    • Haim I Sompolinsky

      The Hebrew University of Jerusalem and Harvard University, Hebrew University of Jerusalem, Center for Brain Science, Harvard Univer

    View abstract →

  • Error-driven Input Modulation: Solving the Credit Assignment Problem without a Backward Pass

    ORAL · Invited

    –

    Publication: G. Dellaferrera, G. Kreiman, Error-driven Input Modulation: Solving the Credit Assignment Problem without a Backward Pass, Manuscript in preparation

    Presenters

    • Giorgia Dellaferrera

      Harvard Medical School and Boston Children's Hospital

    Authors

    • Giorgia Dellaferrera

      Harvard Medical School and Boston Children's Hospital

    • Gabriel Kreiman

      Harvard Medical School and Boston Children's Hospital

    View abstract →

  • Dynamical phases and computation in nonlinear networks with correlated couplings

    ORAL

    –

    Publication: D. Wennberg, S. Ganguli, and H. Mabuchi. Spectra of matrices with partially symmetric randomness. Forthcoming.<br>D. Wennberg, A. Yamamura, S. Ganguli, and H. Mabuchi. Forthcoming.

    Presenters

    • Daniel Wennberg

      Stanford University

    Authors

    • Daniel Wennberg

      Stanford University

    • Atsushi Yamamura

      Stanford University

    • Surya Ganguli

      Stanford, Stanford University

    • Hideo Mabuchi

      Stanford University

    View abstract →

  • Capacity of Group-invariant Linear Readouts from Equivariant Representations: How Many Objects can be Linearly Classified Under All Possible Views?

    ORAL

    –

    Publication: arXiv (https://arxiv.org/abs/2110.07472).<br>Submitted to ICLR 2022 (https://iclr.cc).

    Presenters

    • Matthew S Farrell

      Harvard University

    Authors

    • Matthew S Farrell

      Harvard University

    • Blake Bordelon

      Harvard University

    • Shubhendu Trivedi

      Massachusetts Institute of Technology

    • Cengiz Pehlevan

      Harvard University

    View abstract →

  • Signal representation and learning in random feedback neural networks

    ORAL

    –

    Publication: Susman, L., Mastrogiuseppe, F., Brenner, N., & Barak, O. (2021). Quality of internal representation shapes learning performance in feedback neural networks. Physical Review Research, 3(1), 013176.‏

    Presenters

    • Lee Susman

      Princeton University

    Authors

    • Lee Susman

      Princeton University

    • Francesca Mastrogiuseppe

      University College London

    • Naama Brenner

      Technion Israel Institute of Technology, Technion - Israel Institute of Technolog

    • Omri Barak

      Technion Israel Institute of Technology

    View abstract →

  • Understanding multi-pass stochastic gradient descent via dynamical mean-field theory

    ORAL

    –

    Publication: - The effective noise of stochastic gradient descent and how local knowledge of partial information drives complex systems, Francesca Mignacco, Pierfrancesco Urbani, Article in preparation.<br><br>- Stochasticity helps to navigate rough landscapes: comparing gradient-descent-based algorithms in the phase retrieval problem, Francesca Mignacco, Pierfrancesco Urbani, Lenka Zdeborova, Machine Learning: Science and Technology, 2021.<br><br>- Dynamical mean-field theory for stochastic gradient descent in Gaussian mixture classification, Francesca Mignacco, Florent Krzakala, Pierfrancesco Urbani and Lenka Zdeborova, Advances in Neural Information Processing Systems, 2020, vol. 33.<br>To appear in the "Machine Learning 2021'' Special Issue, JSTAT.

    Presenters

    • Francesca Mignacco

      Institute of Theoretical Physics, CEA Saclay

    Authors

    • Francesca Mignacco

      Institute of Theoretical Physics, CEA Saclay

    View abstract →