Code for Deep Convolutional Neural Network Compression via Coupled Tensor Decomposition

Feel free to ask me questions, and please cite our work if it help:

@ARTICLE{9261106,
  author={Sun, Weize and Chen, Shaowu and Huang, Lei and So, Hing Cheung and Xie, Min},
  journal={IEEE Journal of Selected Topics in Signal Processing}, 
  title={Deep Convolutional Neural Network Compression via Coupled Tensor Decomposition}, 
  year={2021},
  volume={15},
  number={3},
  pages={603-616},
  doi={10.1109/JSTSP.2020.3038227}}

---

ISTA+

Baseline model

The Baseline model and training data is from the paper "ISTA-Net: Interpretable Optimization-Inspired Deep Network for Image Compressive Sensing".

Run the 'Train_Code_for_ISTA_Net_plus.py' to train the original ISTA plus net with parameter sharing and get ' para_dict.npy ' file which including the parameters.

TT-compression

'ISTA_Net_TT_compression.py' uses the independent TT decomposition.

NC-CTD

'5_25_joint_TT_compression.py' uses the NC-CTD algorithm.

Resnet18

Baseline model

Run the 'resnet18.py' using the cifar10 dataset to get the weights parameters file 'parameter.npy'

Single compression

The independent TT and SVD compression methods are showed in 'resnet18_single_tt.py' and 'resnet18_svd.py' respectively.

NC-PCTD

'algrithm1_para_npy' includes the weights by using the algrithm1, and 'algrithm2_npy' includes the weights by using the algrithm2.

'joint_net.py' uses the NC-PCTD approach.