Birs- 24w5301: Structured Machine Learning and Time–Stepping for Dynamical Systems

  1. Title:
    Dynamical systems in deep generative modelling

    Speaker:

    Link:

  2. Title:
    Machine learning of conservation laws for dynamical systems

    Speaker:

    Link:

  3. Title:
    Learning Lagrangian dynamics from data with UQ

    Speaker:

    Link:

  4. Title:
    Stability of numerical methods set on Euclidean spaces and manifolds with applications to neural networks

    Speaker:

    Link:

  5. Title:
    Practical existence theorems for deep learning approximation in high dimensions

    Speaker:

    Link:

  6. Title:
    Improving the robustness of Graph Neural Networks with coupled dynamical systems

    Speaker:

    Link:

  7. Title:
    Time dependent graph neural networks

    Speaker:

    Link:

  8. Title:
    Conservative Hamiltonian Monte Carlo

    Speaker:

    Link:

  9. Title:
    (Lie-group) Structured Inverse-free Second-order Optimization for Large Neural Nets

    Speaker:

    Link:

  10. Title:
    Optimization and Sampling in Non-Euclidean Spaces

    Speaker:

    Link:

  11. Title:
    The Connections Between Discrete Geometric Mechanics, Information Geometry, Accelerated Optimization and Machine Learning

    Speaker:

    Link:

  12. Title:
    Deep neural networks on diffeomorphism groups for optimal shape reparameterization

    Speaker:

    Link:

  13. Title:
    Control and neural network uncertainty quantification for plasma simulation

    Speaker:

    Link:

  14. Title:
    Adaptivity and expressivity in neural network approximations

    Speaker:

    Link:

  15. Title:
    Efficient gradient descent algorithms for learning from multiscale data

    Speaker:

    Link:

  16. Title:
    Data-driven modeling of complex chaotic dynamics on invariant manifolds

    Speaker:

    Link:

  17. Title:
    A spatiotemporal discretization for diffeomorphism approximation

    Speaker:

    Link:

  18. Title:
    A particle method based on Voronoi decomposition for the Cahn–Hilliard equation

    Speaker:

    Link:

  19. Title:
    Explicit time discretizations that preserve dissipative or conservative energy dynamics

    Speaker:

    Link:

  20. Title:
    Geometry aware neural operators for hemodynamics

    Speaker:

    Link: