JSM2023
Back to the program
Topic-Contributed Paper Session

Empirically Driven Deep Learning Theory

Wed, Aug 9, 10:30 AM
Vardan PapyanChairWeijie SuOrganizerX.Y. HanOrganizer
SSC (Statistical Society of Canada)

5 Presentations

A Law of Data Separation in Deep Learning
10:35 AM - 10:55 AM
Hangfeng He (University of Rochester)
Classification versus regression in overparameterized regimes: Does the loss function matter?
10:55 AM - 11:15 AM
Vidya Muthukumar (Georgia Institute of Technology)
Feature Learning in Two-layer Neural Networks
11:15 AM - 11:35 AM
Murat Erdogdu (University of Toronto)
Neural Collapse in Deep Net Training
11:35 AM - 11:55 AM
X.Y. Han (Cornell University)
On the Role of Neural Collapse in Transfer and Few-Shot Learning
11:55 AM - 12:15 PM
Tomer Galanti
↑