模型压缩备用

categorized into four schemes:

• parameter pruning and sharing
• low-rank factorization
• transferred/compact convolutional filters
• knowledge distillation

Achieving compacting and accelerating CNNs model calls for joint solutions from many disciplines, including but not limited to machine learning, optimization, computer architecture, data compression, indexing,and hardware design, compressing and accelerating deep neural networks

parameter pruning and sharing

can be further classified into three sub-categories:

• quantization and binarization
• Pruning and Sharing
• structural matrix.

Quantization and Binarization

• Network quantization compresses the original network by
  reducing the number of bits required to represent each weight

Low-rank factorization

Transferred/compact convolutional filters

knowledge distillation (KD)

• Improved Knowledge Distillation via Teacher Assistant