Class-dependent Compression of Deep Neural Networks

Rahim Entezari, Olga Saukh

Today's deep neural networks require substantial computation resources for their training, storage, and inference, which limits their effective use on resource-constrained devices.…