Diffusion-Jump-GNNs

Diffusion-Jump GNNs: Homophiliation via Learnable Metric Filters

Dependencies

Conda environment

conda create --name <env> --file requirements.txt

or

conda env create -f conda_graphy_environment.yml
conda activate graphy

Code organization

• splits/: splits that we used, taking from GEO-GCN repository.
• main.py: script with inline arguments for running the experiments.
• main_scalable.py: script with inline arguments for running large experiments.
• models.py: script with our proposed architecture.
• pump.py: implementation of the proposed pump.
• utils.py: extra functions used for the experiments.

Run experiments

python main.py --dataset texas --hidden_channels 64 --dropout 0.2 --lr 0.03 --n_layers 20 --epochs 700 --cuda cpu
python main.py --dataset wisconsin --hidden_channels 64 --dropout 0.5 --lr 0.03 --n_layers 5 --epochs 700 --cuda cpu
python main.py --dataset cornell --hidden_channels 128 --dropout 0.5 --lr 0.03 --n_layers 5 --wd 0.001 --epochs 700 --cuda cpu