Pruning deep neural networks to make them fast and small

My PyTorch implementation of [1611.06440 Pruning Convolutional Neural Networks for Resource Efficient Inference]. TL;DR: By using pruning a VGG-16 based Dogs-vs-Cats classifier is made x3 faster and x4 smaller. Pruning neural networks is an old idea going back to 1990 (with Yan Lecun’s optimal brain damage work) and before. The idea is that among the many parameters in the network, some are redund