Proxy Convexity: A Unified Framework for the Analysis of Neural Networks Trained by Gradient Descent

Proxy Convexity: A Unified Framework for the Analysis of Neural Networks Trained by Gradient Descent

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


Duration: 10:46
115 views
0

Spencer Frei (UC Berkeley)
Meet the Fellows Welcome Event



Other Videos By Simons Institute for the Theory of Computing


2021-09-13Understanding Statistical-to-Computational Gaps via Low-Degree Polynomials
2021-09-13Sharp Threshold for the Planted Matching Problem
2021-09-13Correlation Adjusted Debiasing (CAD): Debiasing the Lasso with Inaccurate Covariate Model
2021-09-13On Query-to-Communication Lifting for Adversary Bounds
2021-09-13One-Way Functions Imply Secure Computation in a Quantum World
2021-09-13Quantum Algorithms for the Mean Estimation Problem
2021-09-13Quantum Information Meets Cryptography
2021-09-13Distribution-Free, Risk-Controlling Prediction Sets
2021-09-13Learning Under Requirements
2021-09-13Deep Neural Tangent Kernel and Laplace Kernel Have the Same RKHS
2021-09-13Proxy Convexity: A Unified Framework for the Analysis of Neural Networks Trained by Gradient Descent
2021-09-13On Implicit Regularization in Deep Learning
2021-09-13Equilibrium Computation and the Foundations of Multi-Agent Machine Learning
2021-09-13A Fast Algorithm for Optimal Transport
2021-09-13Approximating Distributions Using Well-Conditioned Normalizing Flows
2021-09-13Iterative Methods and High-Dimensional Statistics
2021-09-13Geometric Methods for Machine Learning and Optimization
2021-09-13Optimal Transport for Inverse Problems
2021-09-13Graphical Optimal Transport and Its Applications
2021-09-13Modularity and Edge-Sampling
2021-09-13The Seeds of Stories


Tags:
Simons Institute
theoretical computer science
UC Berkeley
Computer Science
Theory of Computation
Theory of Computing
Meet the Fellows Welcome Event
Spencer Frei