Introduction

This repository contains the source code for the publications [From Coupled Oscillators to Graph Neural Networks: Reducing Over-smoothing via a Kuramoto Model-based Approach].

The source code is modified from the code of Graph Neural Diffusion (GRAND) (https://github.com/twitter-research/graph-neural-pde). Please see the GRAND's repo in case of environment setting and troubleshooting.

Running the experiments

A general command to run the code:

python3 run_GNN.py --method dopri5 --time $time --no_early --coupling_strength $coup --dataset $data

For example, for the Citeseer dataset:

python3 run_GNN.py --method dopri5 --time 16 --no_early --coupling_strength 0.9 --dataset Citeseer

Requirements

Dependencies (with python >= 3.7): Main dependencies are torch==1.8.1 torch-cluster==1.5.9 torch-geometric==1.7.0 torch-scatter==2.0.6 torch-sparse==0.6.9 torch-spline-conv==1.2.1 torchdiffeq==0.2.1 Commands to install all the dependencies in a new conda environment

conda create --name grand python=3.7
conda activate grand

pip install ogb pykeops
pip install torch==1.8.1
pip install torchdiffeq -f https://pytorch-geometric.com/whl/torch-1.8.1+cu102.html

pip install torch-scatter -f https://pytorch-geometric.com/whl/torch-1.8.1+cu102.html
pip install torch-sparse -f https://pytorch-geometric.com/whl/torch-1.8.1+cu102.html
pip install torch-cluster -f https://pytorch-geometric.com/whl/torch-1.8.1+cu102.html
pip install torch-spline-conv -f https://pytorch-geometric.com/whl/torch-1.8.1+cu102.html
pip install torch-geometric