Zusammenfassung für den Vortrag am 11.02.2021 (17:00 Uhr)Math Machine Learning seminar MPI MIS + UCLA
Samuel L. Smith (DeepMind)
A Backward Error Analysis for Random Shuffling SGD
Siehe auch das Video dieses Vortrages.
I will discuss two recent works, which apply backward error analysis to describe the influence of finite learning rates on the dynamics of Gradient Descent  and Random Shuffling SGD . In particular, I will show that, when using small finite learning rates, the path taken by Random Shuffling SGD coincides with the path taken by gradient flow when minimizing a modified loss function. This modified loss function contains an implicit source of regularization which enhances the test accuracies achieved by deep networks on standard benchmarks.
I will also briefly discuss an empirical study clarifying the key phenomena observed when training deep networks with SGD .
 Implicit Gradient Regularization, Barrett and Dherin, ICLR 2021
 On the Origin of Implicit Regularization in Stochastic Gradient Descent, Smith et al., ICLR 2021
 On the Generalization Benefit of Noise in Stochastic Gradient Descent, Smith et al., ICML 2020