Fetching the paper…

The Law of Parsimony in Gradient Descent for Learning Deep Linear Networks · Around