no code implementations • 8 Jul 2020 • Samuel Glass, Simeon Spasov, Pietro Liò
A novel method to identify salient computational paths within randomly wired neural networks before training is proposed.
