Adaptive Message Passing (AMP)

This is the official code to reproduce the experiments of our JMLR submission

Adaptive Message Passing: A General Framework to Mitigate Oversmoothing, Oversquashing, and Underreaching

Authors:

• Federico Errica
• Henrik Christiansen
• Viktor Zaverkin
• Takashi Maruyama
• Mathias Niepert
• Francesco Alesiani

Citing us

@article{errica2023adaptive,
  title={Adaptive Message Passing: A General Framework to Mitigate Oversmoothing, Oversquashing, and Underreaching},
  author={Errica, Federico and Christiansen, Henrik and Zaverkin, Viktor and Maruyama, Takashi and Niepert, Mathias and Alesiani, Francesco},
  journal={arXiv preprint arXiv:2312.16560},
  year={2023}
}

How to reproduce our experiments

Install all requirements (Python 3.10)

Create a new virtual environment with Python 3.10 installed. Then

# install torch 2.0.1 among other dependencies
pip install -r requirements.txt

# install specific version of PyG to ensure no changes can influence results
pip install torch_geometric==2.3.1

# install ther dependencies (**CHANGE cu117 with your CUDA version**):
pip install torch_scatter==2.1.1 torch_sparse==0.6.17 -f https://data.pyg.org/whl/torch-2.0.1+cu117.html

Data Preparation

Run the notebook Data_Converter.ipynb to prepare the datasets for the experiments.

Launch experiments (assumes CUDA is used)

source launch_synthetic_300patience.sh
source launch_peptides.sh

Results will be stored in the RESULTS folder according to the PyDGN format.

Troubleshooting

Please open an issue if you have trouble running the experiments.

Running on CPU

If you need to run on CPU, a few minor changes need to be made in the configuration files of the experiments. In particular, use

# Hardware
device:  cuda
max_cpus:  [NUM CPUs you want to use]
max_gpus: 0
gpus_per_task:  0

# Data Loading
dataset_getter: pydgn.data.provider.DataProvider
data_loader:
  class_name: torch_geometric.loader.DataLoader
  args:
    num_workers : 0
    pin_memory: False

Please refer to PyDGN's documentation.