Awesome

admm-pruning

Prune DNN using Alternating Direction Method of Multipliers (ADMM)

Our paper

“A Systematic DNN Weight Pruning Framework using Alternating Direction Method of Multipliers” (https://arxiv.org/abs/1804.03294)

Citation

If you use these models in your research, please cite:

@article{zhang2018systematic,
  title={A Systematic DNN Weight Pruning Framework using Alternating Direction Method of Multipliers},
  author={Zhang, Tianyun and Ye, Shaokai and Zhang, Kaiqi and Tang, Jian and Wen, Wujie and Fardad, Makan and Wang, Yanzhi},
  journal={arXiv preprint arXiv:1804.03294},
  year={2018}
}

Models

1. lenet-5

• see tensorflow-mnist-model in this repository

2. bvlc_alexnet (focus on weight reduction)

• [bvlc_alexnet_21x_total.caffemodel] (http://bit.ly/2JVwKFD)

3. bvlc_alexnet (focus on conv reduction)

• [bvlc_alexnet_13_4x_conv.caffemodel] (http://bit.ly/2uNotPv)

Results

1. lenet-5 (top1 accuracy: 99.2%)

2. bvlc_alexnet (top5 accuracy: 80.2%, 40 iterations of ADMM)