Gradient descent provably escapes saddle points in the training of shallow ReLU networks

Add code
Aug 03, 2022

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: