NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:3315
Title:Learning and Generalization in Overparameterized Neural Networks, Going Beyond Two Layers

This paper presents analysis of what 3-layer neural nets with non-smooth activation functions can learn. Extending such analysis to three layers is very challenging and all reviewers find this result interesting. Therefore, I recommend acceptance.