Reviews: Discrimination-aware Channel Pruning for Deep Neural Networks

The paper presents an approach for channel pruning in deep neural networks. The authors propose to prune channels based on their "discriminative power" -- this is achieved by introducing additional per-layer losses, and performing channel selection and network parameter optimization in an iterative fashion. Experimental results on 3 datasets (CIFAR-10, ILSVRC-12, LFW) show that the method compares favorably with respect to the state of the art. Overall, the paper is well written, and explanations are clear. Material provides many further details / experiments.