Birs- 24w5297: Mathematics of Deep Learning

  1. Title:
    Yes, my deep network works! But.. what did it learn?

    Speaker:

    Link:

  2. Title:
    Lecture I: White-Box Architecture Design via Unrolled Optimization and Compression

    Speaker:

    Link:

  3. Title:
    Equivariant convolutional networks over arbitrary Lie groups

    Speaker:

    Link:

  4. Title:
    A Precise High-Dimensional Statistical Theory for Convex and Nonconvex Matrix Sensing

    Speaker:

    Link:

  5. Title:
    Constrained Reinforcement Learning

    Speaker:

    Link:

  6. Title:
    Surprising phenomena of max-ℓ

    Speaker:

    Link:

  7. Title:
    Learning Dynamics of Overparametrized Neural Networks

    Speaker:

    Link:

  8. Title:
    Probably Approximately Correct in the Future.

    Speaker:

    Link:

  9. Title:
    Gradient Descent and Attention Models: Challenges Posed by the Softmax Function

    Speaker:

    Link:

  10. Title:
    Where within F_1 does gradient descent search?

    Speaker:

    Link:

  11. Title:
    Data Reconstruction Attacks and Defenses: From Theory to Practice

    Speaker:

    Link:

  12. Title:
    A Poincaré Inequality and Consistency Results for Signal Sampling on Large Graphs

    Speaker:

    Link:

  13. Title:
    The Role of Sparsity in Differentially-Private Learning

    Speaker:

    Link:

  14. Title:
    Symmetries in machine learning: point clouds and graphs.

    Speaker:

    Link:

  15. Title:
    How many FLOPs is a token worth?

    Speaker:

    Link:

  16. Title:
    Lazy quotient metrics: Approximate symmetries for ML models

    Speaker:

    Link:

  17. Title:
    Robust Unrolled Networks

    Speaker:

    Link:

  18. Title:
    Overcoming the Boundaries of Artificial Intelligence: A Mathematical Approach

    Speaker:

    Link: