Heterogeneous Graph Neural Network via Attribute Completion

This repository contains the demo code of the paper:

Heterogeneous Graph Neural Network via Attribute Completion

which has been accepted by WWW2021.

Dependencies

• Python3
• NumPy
• SciPy
• scikit-learn
• NetworkX
• DGL
• PyTorch

Datasets

The preprocessed datasets are available at Baidu Netdisk(password: hgnn) or Google Drive.

Please extract the zip file to folder data.

We use the same methods as MAGNN to process the data, so you can also download datasets at MAGNN's repository.

Example

• python run_DBLP.py --cuda
• python run_IMDB.py --cuda
• python run_ACM.py --cuda

Please refer to the code for more parameters.

Acknowledgements

The demo code is implemented based on MAGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding.

Citing

@inproceedings{hgnn-ac,
 title={Heterogeneous Graph Neural Network via Attribute Completion},
 author={Di Jin and Cuiying Huo and Chundong Liang and Liang Yang},
 booktitle = {WWW},
 year={2021}
}