Fetching the paper…

How many degrees of freedom do we need to train deep networks: a loss landscape perspective · Around