Data Augmentation for Graph Neural Networks

This repository contains the source code for the AAAI'2021 paper:

Data Augmentation for Graph Neural Networks

by Tong Zhao ([email protected]), Yozen Liu, Leonardo Neves, Oliver Woodford, Meng Jiang, and Neil Shah.

Requirements

This code package was developed and tested with Python 3.7.6. Make sure all dependencies specified in the requirements.txt file are satisfied before running the model. This can be achieved by

pip install -r requirements.txt

Usage

The scripts for hyperparameter search with Optuna are optuna_[method].py.

All the parameters are included in best_parameters.json. Results can be reproduced with the scripts train_[method].py, which will automatically load the parameters. For example, to reproduce the result of GAugO with GCN on Cora, you can simply run:

python train_GAugO.py --dataset cora --gnn gcn --gpu 0

Data

The format of data files are described in detail in the file data/README. Due to file size limit, for GAugM, only the edge_probabilities of Cora is provided. Please find the all edge_probabilities files at https://tinyurl.com/gaug-data. The VGAE implementation I used for generating these edge_probabilities are also provided under the folder vgae/.

Cite

If you find this repository useful in your research, please cite our paper:

@inproceedings{zhao2021data,
  title={Data Augmentation for Graph Neural Networks},
  author={Zhao, Tong and Liu, Yozen and Neves, Leonardo and Woodford, Oliver and Jiang, Meng and Shah, Neil},
  booktitle={The Thirty-Fifth AAAI Conference on Artificial Intelligence},
  pages={},
  year={2021}
}