BiFormer

The source code of BiFormer: A Framework Fusing Transformer and Graph Convolution for Large-Scale Graphs

Training

To reproduce the results of BiFormer on the Cora dataset, please run following commands.

Conda activate your_conda_env_name
python test_Cora.py 

If you need to apply BiFormer to other datasets, simply modify the dataset section and use the hyperparameters suggested in the paper for training.

Citing

Please cite our work if you find it is useful for you:

@article{BiFormerBipartiteStreamInformation,
  title = {BiFormer: A Bipartite-Stream Information Fusion Framework for Large-Scale Graph Representation Learning},
  author = {Zhang, Qi and Sun, Yanfeng and Wang, Shaofan and Gao, Junbin and Yin, Baocai},
  year = 2025,
  journal = {Neural Networks},
  volume = {191},
  pages = {107792}
}