When Both Layers Learn: Training Dynamics of Representing Linear Models via ReLU Networks.

Berk Tinaz, Changzhi Xie, Mahdi Soltanolkotabi

Browse the full COLT paper archive.