NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:9118
Title:Are deep ResNets provably better than linear predictors?

This paper builds upon and extends the work of Shamir'18 in an attempt to study the power of Resnets. They provide several motivating examples and new results regarding deep resnets. While the results here seem to rely on some strong assumptions, overall it seems that the authors are able to make a certain progress using non trivial techniques. Hopefully future work will either show these assumption to be redundant or necessary.