Improving Expressivity of GNNs with Subgraph-specific Factor Embedded Normalization (KDD'23)

Overview

Official code for paper Improving Expressivity of GNNs with Subgraph-specific Factor Embedde Normalization, which proposes a dedicated plug-and-play normalization scheme to strengthen the representative capabilities of GNNs, that is:

extending GNNs to be at least as powerful as the 1-WL test in distinguishing non-isomorphism graphs.
alleviating over-smoothing issue while GNNs' layers going deeper.

Contact: Please feel free to contact me via email (chenkx@zju.edu.cn) if you have any questions.

Citation: If you find this repository helpful for your research, please kindly cite the following paper:

@inproceedings{chen2023improving,
  title={Improving Expressivity of GNNs with Subgraph-specific Factor Embedded Normalization},
  author={Chen, Kaixuan and Liu, Shunyu and Zhu, Tongtian and Qiao, Ji and Su, Yun and Tian, Yingjie and others},
  booktitle={ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD)},
  year={2023}
}

1. Environment Installation

conda create -n torch python=3.9
conda activate torch
conda install pytorch==1.10.0 torchvision cudatoolkit=11.3 -c pytorch
pip install dgl-cu113==0.7.1 -f https://data.dgl.ai/wheels/repo.html
pip install ogb
pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-1.10.0+cu113.html

2. Usage

2.1 quick start on ogbg-moltoxcast

download dataset

python download_dataset.py --dataset_name 'ogbg-moltoxcast'

run 4 layers GCN using supernorm and batchnorm

python main_graph.py --model 'GCN' --num_layer 4 --norm_type 'supernorm'
python main_graph.py --model 'GCN' --num_layer 4 --norm_type 'batchnorm'

run 16 layers GCN using supernorm and batchnorm

python main_graph.py --model 'GCN' --num_layer 16 --norm_type 'supernorm'
python main_graph.py --model 'GCN' --num_layer 16 --norm_type 'batchnorm'

2.2 Experiments for Graph Isomorphism Test

Firstly, download dataset:

python download_dataset.py --dataset_name 'imdb-binary'

Then, please find shell files in 'cripts/graph-imdb-sl', and running

sh scripts/graph-imdb-sl/run_gin_sl_supernorm.sh

To reproduce the Figure 3 in the paper, please remove the warmup operation (i.e, delete --lr_warmup in shell files.)

2.3 Experiments for Over-smoothing Issue

Firstly, download dataset:

python download_dataset.py --dataset_name 'cora'

Then, please find shell files in 'scripts/node-cora', and running

sh scripts/node-cora/run_gcn_supernorm.sh

2.4 Experiments on other datasets

2.4.1 Graph-Level:

download ogbg-moltoxcast:

   python download_dataset.py --dataset_name 'ogbg-moltoxcast'

Then, please find shell files in 'scripts/ogbg-toxcast', and running

   sh scripts/ogbg-toxcast/run_gcn_supernorm.sh

2.4.2 Node-Level:

download pumbed:

   python download_dataset.py --dataset_name 'pubmed'

Then, please find shell files in 'scripts/node-pubmed', and running

   sh scripts/node-pubmed/run_gcn_supernorm.sh

2.4.3 Link-Level:

download ogbl-collab:

   python download_dataset.py --dataset_name 'ogbl-collab'

Then, please find shell files in 'scripts/ogbl-collab', and running

   sh scripts/ogbl-collab/run_gcn_supernorm.sh
   sh scripts/ogbl-collab/run_gcn_batchnorm.sh

chenchkx / SuperNorm