HGCN (Hierarchical Graph Capsule Network)

Code of the AAAI 2021 paper HGCN --- Supplementary

Prerequisites and Dependencies

Option 1: Nvidia Docker image

docker pull utasmile/pytorch:1.4.0-cuda10.0-cudnn7.5-devel

Option 2: Install the dependencies by yourself

• Pytorch >= 1.4.0
• Pytorch Geometric

Data

All the datasets used in the paper can be downloaded from Benchmark Data Sets for Graph Kernels

Model Training

All the commands can be found in script.txt For example:

python main.py --epochs 100 --batch_size 256 --capsule-dimensions 128 --capsule-num 10 --dropout 0.1 --theta 0.1 --lr 0.003 --dataset IMDB-BINARY

Citation

If you use HGCN in your research, please cite the following paper:

@article{yang2020hierarchical,
  title={Hierarchical Graph Capsule Network},
  author={Yang, Jinyu and Zhao, Peilin and Rong, Yu and Yan, Chaochao and Li, Chunyuan and Ma, Hehuan and Huang, Junzhou},
  journal={AAAI Conference on Artificial Intelligence (AAAI)},
  year={2021}
}

The code is largely borrowed from:

@article{chen2019powerful,
  title={Are powerful graph neural nets necessary? a dissection on graph classification},
  author={Chen, Ting and Bian, Song and Sun, Yizhou},
  journal={arXiv preprint arXiv:1905.04579},
  year={2019}
}