Understanding and Improving Generalization in Deep Learning

Schedule

8:30 -8:40 Opening Remarks

8:40 - 9:20 Invited Speaker: Daniel Roy "Progress on Nonvacuous Generalization Bounds"

9:20 - 9:50 Invited Speaker: Chelsea Finn "Training for Generalization"

9:50 - 10:05 Spotlight Talk: "A Meta-Analysis of Overfitting in Machine Learning"

10:05 - 10:20 Spotlight Talk: "Uniform convergence may be unable to explain generalization in deep learning"

10:20 - 10:40 Break and Poster Session

10:40 - 11:10 Invited Speaker: Sham Kakade "Prediction, Learning, and Memory"

11:10 - 11:40 Invited Speaker: Mikhail Belkin "A Hard Look at Generalization and its Theories"

11:40 - 11:55 Spotlight Talk: "Towards Task and Architecture-Independent Generalization Gap Predictors"

11:55 - 12:10 Spotlight Talk: "Data-Dependent Sample Complexity of Deep Neural Networks via Lipschitz Augmentation"

12:10 - 13:30 Lunch and Poster Session

13:30 - 14:00 Invited Speaker: Aleksander Mądry "Are All Features Created Equal?"

14:00 - 14:30 Invited Speaker: Jason Lee "On the Foundations of Deep Learning: SGD, Overparametrization, and Generalization"

14:30 - 14:45 Spotlight Talk: "Towards Large Scale Structure of the Loss Landscape of Neural Networks"

14:45 - 15:00 Spotlight Talk: "Zero-Shot Learning from scratch: leveraging local compositional representations"

15:00 - 15:30 Break and Poster Session

15:30 - 16:30 Panel Discussion (Moderator: Nati Srebro)

16:30 - 16:45 Spotlight Talk: "Overparameterization without Overfitting: Jacobian-based Generalization Guarantees for Neural Networks"

16:45 - 17:00 Spotlight Talk: "How Learning Rate and Delay Affect Minima Selection in Asynchronous Training of Neural Networks: Toward Closing the Generalization Gap"

17:00 - 18:00 Poster Session