|
|
|
|
283
|
Wed, 8/5/2020,
10:00 AM -
11:50 AM
|
Virtual
|
|
Theoretical Advances in Deep Learning — Invited Papers
|
|
IMS, Section on Nonparametric Statistics, International Indian Statistical Association
|
|
Organizer(s): Po-Ling Loh, UW-Madison
|
|
Chair(s): Po-Ling Loh, UW-Madison
|
|
10:05 AM
|
Good Linear Classifiers Are Abundant in the Interpolating Regime
Jason Klusowski, Rutgers University; Ryan Theisen, UC Berkeley; Michael Mahoney, UC Berkeley
|
|
10:20 AM
|
Multiple Descent Phenomenon, Risk of Minimum-Norm Interpolants and Restricted Lower Isometry of Kernels
Tengyuan Liang, University of Chicago
|
|
10:35 AM
|
Polylogarithmic Width Suffices for Gradient Descent to Achieve Arbitrarily Small Test Error with Shallow ReLU Networks
Matus Telgarsky, University of Illinois - Urbana Champaign
|
|
10:50 AM
|
Discussant: Edgar Dobriban, University of Pennsylvania
|
|
11:05 AM
|
Discussant: Daniel Hsu, Columbia University
|
|
11:20 AM
|
Discussant: Mikhail Belkin, Ohio State University
|
|
11:30 AM
|
Floor Discussion
|
