Deep Neural Network Compression Method Based on Product Quantization | IEEE Conference Publication | IEEE Xplore