bywmm / Bi-GCN

Implementation of "Binary Graph Convolutional Network", CVPR 2021, and TPAMI 2024.

Bi-GCN

Official Implementation of CVPR 2021 Paper: [Bi-GCN: Binary Graph Convolutional Network](https://arxiv.org/abs/2010.07565, and TPAMI 2024 Paper: Binary Graph Convolutional Network With Capacity Exploration.

Please cite our paper if you use this code in your own work:

@INPROCEEDINGS{wang2021,
  author={Wang, Junfu and Wang, Yunhong and Yang, Zhen and Yang, Liang and Guo, Yuanfang},
  booktitle={2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, 
  title={Bi-GCN: Binary Graph Convolutional Network}, 
  year={2021},
  pages={1561-1570}

@ARTICLE{wang2024,
  author={Wang, Junfu and Guo, Yuanfang and Yang, Liang and Wang, Yunhong},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
  title={Binary Graph Convolutional Network With Capacity Exploration}, 
  year={2024},
  volume={46},
  number={5},
  pages={3031-3046}

Requirements

• torch==1.7.0
• torch_geometric==1.7.0
• scikit_learn

Run

Run the demo of Bi-GCN on Cora dataset by this command.

python transductive-bigcn.py --device 0

You can specify a dataset, set the layer number, or other hyper-parameters by setting the optional args.

python bi-gcn.py --gpu 0 --dataset Cora --layers 4

You can run the file inductive-gs-bignn.py and inductive-ns-bignn.py to get the results of binarized version of other GNNs, like inductive GCN, GraphSAGE, and GraphSAINT.

python inductive-ns-bignn.py --device 6 --model GraphSAGE --dataset Reddit --binarize

The shell script of the reported results in Table 2, 3 can be found in results.sh.