DEEP COMPRESSION: COMPRESSING DEEP NEURAL NETWORKSWITH PRUNINGTRAINED QUANTIZATION ANDHUFFMANCODING

DEEP COMPRESSION: COMPRESSING DEEP NEURAL NETWORKS WITH PRUNING, TRAINED QUANTIZATION AND HUFFMAN CODING: 深度压缩：用剪枝、训练量化和霍夫曼编码压缩深度神经网络

• 第一篇
• ABSTRACT
• 1 INTRODUCTION
• 2 NETWORK PRUNING(网络剪枝-暂时不看这部分！
• 3 TRAINED QUANTIZATION AND WEIGHT SHARING(量化和分享训练有素的权重)
• • 3.1 WEIGHT Sharing
  • 3.2 INITIALIZATION OF SHARED WEIGHTS(初始权重)
  • 3.3 FEED-FORWARD AND BACK-PROPAGATION(前向传播和反向传播)
• 4 HUFFMAN CODING（霍夫曼编码）
• 5 EXPER