On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks

Jiong Zhu, Gaotang Li, Yao-An Yang, Jing Zhu, Xuehao Cui, Danai Koutra. 2024. On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks. Advances in Neural Information Processing Systems 38 (2024).

[Paper]

Run Experiments

To install the dependecies in a new conda environment, run

$ conda create --name <env> --file hetelinkpred/conda_env/dgl.txt

Scripts used to get results for BUDDY can be found in /subgraph-sketching/scripts.
Other scripts can be found in /hetelinkpred/shell-scripts-hetelinkpred.

GNN Encoders Supported

• GraphSAGE
• GCN
• BUDDY

Decoders Supported

• MLP
• DistMult
• Dot product

Heuristics Supported

• Common Neighbor
• Adamic Adar
• Resource Allocation
• Personalized Page Rank

Datasets

• Synthetic dataset of varying heterophily level
• ogbl-collab
• ogbl-citation2
• E-Commerce
• Attributed-PPI
• Attributed-Facebook

Please following the script below to unzip the synthetic and e-commerce datasets.

# sudo apt-get install p7zip-full # install 7z for ubuntu.
# brew install p7zip # install 7z for mac
# For windows, download 7z from https://www.7-zip.org/download.html

cd hetelinkped
7z x dataset.7z.part.001
7z x dataset.7z

Other datasets can be downloaded from their repective official sites, and processed with /hetelinkpred/generate_split.py

Contact

Please contact Jiong Zhu (jiongzhu@umich.edu) in case you have any questions.

Citation

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

@article{zhu2024impactfeatureheterophilylink,
    title={On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks}, 
    author={Zhu, Jiong and  Li, Gaotang and Yang, Yao-An and Zhu, Jing and Cui, Xuehao and Koutra, Danai},
    journal={Advances in Neural Information Processing Systems},
    volume={37},
    year={2024}
}