Shrink and Perturb: On Warm-Starting Neural Network Training December 31, 2020 https://arxiv.org/pdf/1910.08475 Fullscreen Dark Mode