Theorinet Retreat 2021
Recordings:
September 20th, 2021: Session 1: Overparameterization
September 20th, 2021 Session 2: Optimization and robustness
September 21st, 2021: Session 1: Learning on graphs
September 21st, 2021: Session 2: Limits of learning
| Time | Speaker | Affiliation | Title of Presentation |
|---|---|---|---|
Monday September 20 |
|||
| Overparameterization | |||
| 012:00-12:30 PM EST | Yaodong Yu | UC Berkeley | Understanding Generalization in Adversarial Training via the Bias-Variance Decomposition |
| 012:30-1:00 PM EST | Salma Tarmoun | Johns Hopkins University | Understanding the Dynamics of Gradient Flow in Overparameterized Linear Models |
| 01:00-1:30 PM EST | Hancheng Min | Johns Hopkins University | On the convergence and implicit bias of overparametrized linear networks |
| 01:30-2:00 PM EST | Teresa Huang | Johns Hopkins University | Dimensionality reduction in overparameterized regression |
| 02:00-2:30 PM EST | Meena Jagadeesan | UC Berkeley | Inductive Bias of Multi-Channel Linear Convolutional Networks with Bounded Weight Norm |
| Optimization and robustness | |||
| 03:30-4:00 PM EST | Yaodong Yu | UC Berkeley | ReduNet: A White-box Deep Network from the Principle of Maximizing Rate Reduction |
| 04:00-4:30 PM EST | Chenwei Wu | Duke University | Guarantees for Tuning the Step Size using a Learning-to-Learn Approach |
| 04:30-5:00 PM EST | Alexander Robey | University of Pennsylvania | Model-Based Robust Deep Learning |
| 05:00-5:30 PM EST | Chinmay Maheshwari | UC Berkeley | Zeroth-Order Methods for Convex-Concave Minmax Problems: Applications to Decision-Dependent Risk Minimization |
Tuesday September 21 |
|||
| Learning on graphs | |||
| 012:00-12:30 PM EST | Sohir Maskey | University of Munich | Transferability of Graph Neural Networks |
| 012:30-1:00 PM EST | Luana Ruiz | University of Pennsylvania | Large-Scale Graph Information Processing |
| 01:00-1:30 PM EST | Alejandro Parada | University of Pennsylvania | Large-Scale Graph Information Processing |
| 01:30-2:00 PM EST | Hans Riess | University of Pennsylvania | GNNs & the Tarski Laplacian |
| Limits of learning | |||
| 03:00-3:30 PM EST | Jayanta Dey | Johns Hopkins University | Lifelong Learning: Theory and Practice |
| 03:30-4:00 PM EST | Anastasios Tsiamis | University of Pennsylvania | Linear Systems can be Hard to Learn |
| 04:00-4:30 PM EST | Lihua Lei | Stanford University | Distribution-Free, Risk-Controlling Prediction Sets |
| 04:30-5:00 PM EST | Natalia Martinez | Duke University | Blind Pareto Fairness and Subgroup Robustness |