How to Train Better: Exploiting the Separability of DNNs

How to Train Better: Exploiting the Separability of DNNs

Channel:
United States Simons Institute for the Theory of Computing
Subscribers:
68,700
Published on ● Video Link: https://www.youtube.com/watch?v=xpM_9OS99M4


Duration: 49:11
479 views
15

Lars Ruthotto (Emory University) and Elizabeth Newman (Emory University)
https://simons.berkeley.edu/talks/how-train-better-exploiting-separability-dnns
Dynamics and Discretization: PDEs, Sampling, and Optimization



Other Videos By Simons Institute for the Theory of Computing


2021-11-03Yes, Generative Models Are The New Sparsity
2021-11-03Panel on Novel Fault-Tolerant Qubits | Quantum Colloquium
2021-11-03On the Power of Differentiable Learning
2021-11-03Simple Architecture for a Fault-Tolerant Qubit | Quantum Colloquium
2021-11-02A Data-Centric View On Robustness
2021-11-02Finding Large Cliques in Random and Semi-Random Graphs
2021-10-30Density Control of Interacting Agent Systems
2021-10-30Discrete-To-Continuum Limits of Dynamical Optimal Transport Problems
2021-10-30Learning Interaction Kernels in Stochastic Systems of...
2021-10-30Mirror Descent Algorithms for Minimizing Interacting Free Energy
2021-10-30How to Train Better: Exploiting the Separability of DNNs
2021-10-29A Particle Method for Degenerate Diffusion, and Applications to Robotic Swarming
2021-10-29Deterministic Particle Approximation for Nonlocal Transport Equations
2021-10-29Extending the JKO Scheme Beyond Wasserstein-2 Gradient Flows
2021-10-29Sampling From Wasserstein Barycenter
2021-10-29Consensus Based Sampling
2021-10-29Continuous Time Stochastic Gradient Descent and Flat Minimum Selection
2021-10-29Mean-Field Theory of Two-Layers Neural Networks: Dimension Free Bounds and Example
2021-10-28Archetypal Analysis
2021-10-28Linearised Optimal Transport Distances
2021-10-28Wasserstein GANs Work Because They Fail (to Approximate the Wasserstein Distance)


Tags:
Simons Institute
theoretical computer science
UC Berkeley
Computer Science
Theory of Computation
Theory of Computing
Dynamics and Discretization: PDEs Sampling and Optimization
Lars Ruthotto