The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks

Stunning evidence for the hypothesis that neural networks work so well because their random initialization almost certainly contains a nearly optim...