This repository provides a refined version of pGNN as described in the ICML'2022 published paper 'p-Laplacian Based Graph Neural Networks'.
The following packages are required:
- pytorch 1.9.0
- torch_geometric 1.7.1
- scikit-learn 0.24.2
- networkx 2.5.1
Run the following command:
$ python main.py --input cora --train_rate 0.025 --val_rate 0.025 --model pgnn --mu 0.1 --p 2 --K 4 --num_hid 16 --lr 0.01 --epochs 1000
Run the following command:
$ bash run_test.sh
If you find the pGNN model useful in your research, please consider citing the original paper:
@inproceedings{DBLP:conf/icml/FuZB22,
author = {Guoji Fu and
Peilin Zhao and
Yatao Bian},
title = {p-Laplacian Based Graph Neural Networks},
booktitle = {International Conference on Machine Learning, {ICML} 2022, 17-23 July
2022, Baltimore, Maryland, {USA}},
series = {Proceedings of Machine Learning Research},
volume = {162},
pages = {6878--6917},
publisher = {{PMLR}},
year = {2022},
url = {https://proceedings.mlr.press/v162/fu22e.html},
}