LW-GNN

An offical PyTorch implementation of "Label-Wise Graph Convolutional Network for Heterophilic Graphs".

Overview

• ./models: This directory contains the proposed LW-GCN.
• ./dataset.py: This is the code to load datasets.
• ./data: The datasets are stored here.
• ./scripts: It contains the scripts to reproduce the major reuslts of our paper.
• ./train_BI.py: The program to train LW-GCN that incorporate the label-wise graph convolution and GCNII with model selection.
• ./train_DIS.py: The program to train LW-GCN\G that removes the model selection module.
• ./log: We provide our training log at ./log as a reference to reproduce the results in the paper

Requirements

python>=3.7
torch==1.7.1
torch-geometric==1.7.2
cuda=11.0

Install Packages

conda install pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=11.0 -c pytorch 
pip install --no-index torch-scatter -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install --no-index torch-sparse -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install --no-index torch-cluster -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install --no-index torch-spline-conv -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install torch-geometric==1.7.1

Experiments

To reproduce the performance in the paper, you can run the bash files in the .\scripts. For example, to get results on cora datasets

# Run the LW-GCN on all datasets
bash train_BI.sh

# Run the LW-GCN\G that does not deploy model selection on all heterophilic graphs
bash train_DIS.sh

You can also check the our recorded log in .\log to help reproduce the experiments we conducted.

Cite

If you find this repo to be useful, please cite our paper. Thank you.

@inproceedings{
dai2022labelwise,
title={Label-Wise Graph Convolutional Network for Heterophilic Graphs},
author={Enyan Dai and Shijie Zhou and Zhimeng Guo and Suhang Wang},
booktitle={Learning on Graphs Conference},
year={2022},
url={https://openreview.net/forum?id=HRmby7yVVuF}
}