Fetching the paper…

Implicit bias of deep linear networks in the large learning rate phase · Around