Birs- 16w5136: Applied Harmonic Analysis, Massive Data Sets, Machine Learning, and Signal Processing

  1. Title:
    Obstacles to AI, Mathematical and Otherwise

    Speaker:

    Link:

  2. Title:
    Fundamental limits and algorithms for recovery under matrix spike models

    Speaker:

    Link:

  3. Title:
    Deep vs. shallow networks

    Speaker:

    Link:

  4. Title:
    Robust Low-rank Modeling for High-dimensional and Massive Datasets

    Speaker:

    Link:

  5. Title:
    Regularized Nonlinear Acceleration

    Speaker:

    Link:

  6. Title:
    Convex (and tractable?) formulations for single hidden layer neural networks

    Speaker:

    Link:

  7. Title:
    Functional Maps and Functional Map Networks for Joint Data Analysis

    Speaker:

    Link:

  8. Title:
    Provable Approximations of Deep Nets

    Speaker:

    Link:

  9. Title:
    Multimodal signal processing using Alternating Diffusion

    Speaker:

    Link:

  10. Title:
    The discrete sign problem: uniqueness, stable recovery and phase retrieval applications

    Speaker:

    Link:

  11. Title:
    Rapid, robust and reliable blind deconvolution via nonconvex optimization

    Speaker:

    Link:

  12. Title:
    Cascaded gain control representations

    Speaker:

    Link:

  13. Title:
    The Deceptive Nature of Generalization in Machine Learning

    Speaker:

    Link:

  14. Title:
    Nonconvex Recovery of Low-Complexity Models

    Speaker:

    Link:

  15. Title:
    Universality laws for randomized dimension reduction

    Speaker:

    Link:

  16. Title:
    Clustering subgaussian mixtures via semidefinite programming

    Speaker:

    Link:

  17. Title:
    Synchronization over Cartan motion groups via contraction

    Speaker:

    Link:

  18. Title:
    Batched Stochastic Gradient Descent with Weighted Sampling

    Speaker:

    Link:

  19. Title:
    Phase Retrieval

    Speaker:

    Link:

  20. Title:
    Simultaneous Alignment and Classification - A Representation Theory Perspective

    Speaker:

    Link: