[Statlist] ETH Young Data Science Researcher Seminar Zurich, Virtual Seminar by Spencer Frei, UCLA

[Maurer Letizia]

Dear all

We are glad to announce the following talk in the virtual ETH Young Data Science Researcher Seminar Zurich

"Provable Generalization of SGD-trained Neural Networks of Any Width in the Presence of Adversarial Label Noise"
by Spencer Frei, UCLA

Time: Friday, 16 April 2021, 16:30 -17:30
Place: Zoom at https://ethz.zoom.us/j/92367940258

Abstract: Can overparameterized neural networks trained by SGD provably generalize when the labels are corrupted with substantial random noise? We answer this question in the affirmative by showing that for a broad class of distributions, one-hidden-layer networks trained by SGD generalize when the distribution is linearly separable but corrupted with adversarial label noise, despite the capacity to overfit. Equivalently, such networks have classification accuracy competitive with that of the best halfspace over the distribution. Our results hold for networks of arbitrary width and for arbitrary initializations of SGD.  In particular, we do not rely upon the approximations to infinite width networks that are typically used in theoretical analyses of SGD-trained neural networks.

M. Azadkia, Y. Chen, G. Chinot, M. L�ffler, A. Taeb

Seminar website: https://math.ethz.ch/sfs/news-and-events/young-data-science.html
[https://ethz.ch/etc/designs/ethz/img/header/eth_default_og.jpg]<https://math.ethz.ch/sfs/news-and-events/young-data-science.html>

Young Data Science Researcher Seminar Zurich � Seminar for Statistics | ETH Zurich<https://math.ethz.ch/sfs/news-and-events/young-data-science.html>
math.ethz.ch



	[[alternative HTML version deleted]]