/LGI-LS

[NeurIPS 2023] Latent Graph Inference with Limited Supervision

Primary LanguageJavaScriptMIT LicenseMIT

LGI-LS (NeurIPS 2023)

Codes for the NeurIPS 2023 paper Latent Graph Inference with Limited Supervision.

Project page

Datasets

The Cora, Citeseer, and Pubmed datasets can be download from here. Please place the downloaded files in the folder data_tf. The ogbn-arxiv dataset will be loaded automatically.

Installation

conda create -n LGI python=3.7.2
conda activate LGI
pip install torch==1.5.1 torchvision==0.6.1
pip install scipy==1.2.1
pip install scikit-learn==0.21.3
pip install dgl-cu102==0.5.2
pip install ogb==1.2.3
wget https://data.pyg.org/whl/torch-1.5.0%2Bcu102/torch_scatter-2.0.5-cp37-cp37m-linux_x86_64.whl
wget https://data.pyg.org/whl/torch-1.5.0%2Bcu102/torch_sparse-0.6.5-cp37-cp37m-linux_x86_64.whl
wget https://data.pyg.org/whl/torch-1.5.0%2Bcu102/torch_cluster-1.5.4-cp37-cp37m-linux_x86_64.whl
wget https://data.pyg.org/whl/torch-1.5.0%2Bcu102/torch_spline_conv-1.2.0-cp37-cp37m-linux_x86_64.whl
pip install torch_scatter-2.0.5-cp37-cp37m-linux_x86_64.whl
pip install torch_sparse-0.6.5-cp37-cp37m-linux_x86_64.whl
pip install torch_cluster-1.5.4-cp37-cp37m-linux_x86_64.whl
pip install torch_spline_conv-1.2.0-cp37-cp37m-linux_x86_64.whl
pip install torch-geometric==1.6.1

Usage

We provide GCN+KNN, GCN+KNN_U, and GCN+KNN_R as examples due to their simplicity and effectiveness. To test their performances on the Pubmed dataset, run the following command:

bash experiments.sh

The experimental results will be saved in the corresponding *.txt file.

Reference

@inproceedings{Jianglin2023LGI,
  title={Latent Graph Inference with Limited Supervision},
  author={Lu, Jianglin and Xu, Yi and Wang, Huan and Bai, Yue and Fu, Yun},
  booktitle={Advances in Neural Information Processing Systems},
  year={2023}
}

@inproceedings{fatemi2021slaps,
  title={SLAPS: Self-Supervision Improves Structure Learning for Graph Neural Networks},
  author={Fatemi, Bahare and Asri, Layla El and Kazemi, Seyed Mehran},
  booktitle={Advances in Neural Information Processing Systems},
  year={2021}
}

Acknowledgement

Our codes are mainly based on SLAPS. For other comparison methods, please refer to their publicly available code repositories. We gratefully thank the authors for their contributions.