Fetching the paper…

Width Provably Matters in Optimization for Deep Linear Neural Networks · Around