Accurate Neural Network Pruning Requires Rethinking Sparse Optimization

Open in new window