Running the experiments

Introduction:

This is a patch version of graph-neural-pde resolve errors while running experiments on local environment. Providing with several visualization notebooks for visualising networks experiment on the project.

Requirements

Dependencies (with python >= 3.8):

Main dependencies are

• torch==1.13.0
• torch-cluster==1.6.0
• torch-geometric==2.2.0
• torch-scatter==2.1.0
• torch-sparse==0.6.15
• torch-spline-conv==1.2.1
• torchdiffeq==0.2.3

Commands to install all the dependencies in a new conda environment

conda env create -f environment.yml 
conda activate grand

pip install ogb pykeops
pip install torch==1.13.0
pip install torchdiffeq

pip install torch-scatter
pip install torch-sparse
pip install torch-cluster
pip install torch-spline-conv
pip install torch-geometric

Troubleshooting

There is a bug in torch-sparse==0.6.15 occur with Apple Silicon chips.

Resolve by running: pip install git+https://github.com/rusty1s/pytorch_sparse.git

GRAND (Graph Neural Diffusion)

Dataset and Preprocessing

create a root level folder

./data

This will be automatically populated the first time each experiment is run.

Experiments

For example to run for Cora with random splits:

cd src
python run_GNN.py --dataset Cora 

Visualization

• Attention Visualization is available on visualise_attention.ipynb
• MNIST Visualization is available on mnist_visualise.ipynb