Skip to content

TaehyeonKim-pyomu/CNN_compression_rank_selection_BayesOpt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

17 Commits
 
 
 
 
 
 

Repository files navigation

BayesOpt-based Global Optimal Rank Selection for Compression of CNNs (Pytorch Implementation)

T. Kim, J. Lee and Y. Choe, "Bayesian Optimization-Based Global Optimal Rank Selection for Compression of Convolutional Neural Networks," in IEEE Access, vol. 8, pp. 17605-17618, 2020, doi: 10.1109/ACCESS.2020.2968357.

Abstract

Finding the optimal rank is a crucial problem beacause the rank is the only hyperparameter for controlling computational complexity and accuracy in compressed CNNs. To solve this problem, we propose a global optimal rank selection method based on Bayesian optimization. By utilizing both a simple objective function and a proper optimization scheme, the proposed method produces a global optimal rank that provides a good trade-off between computational complexity and accuracy degradation.

Usage

  • Decompose a pretrained vgg16 model: python main.py
  • torch: 1.0.0 version
  • tensorly: 0.4.5 version
  • GPyOpt: 1.2.5 version

Reference