Towards Effective 2-bit Quantization: Pareto-optimal Bit Allocation for Deep CNNs Compression
Towards Effective 2-bit Quantization: Pareto-optimal Bit Allocation for Deep CNNs Compression
Zhe Wang,Jie Lin,3 Authors,B. Girod
2019
1 Citations
TLDR
This paper addresses an important problem of how to optimize the bit allocation of weights and activations for deep CNNs compression and proposes a very efficient method to solve the optimization problem via Lagrangian Formulation.