On the loss landscape of a class of deep neural networks with no bad local valleys.

Quynh Nguyen, Mahesh Chandra Mukkamala, Matthias Hein

Browse the full ICLR paper archive.