Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding

Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems with limited hardware resources. To address this limitation, we introduce "