Improving the Effective Receptive Field of Message-Passing Neural Networks [ICML 2025]

Shahaf E. Finder, Ron Shapira Weber, Moshe Eliasof, Oren Freifeld and Eran Treister

Updates

• June 11, 2025 - Added City-networks training script, and GatedGCNConv option.

How to use

Example:

from models.immpnn import IMMPNN

model = IMMPNN(
  in_channels=37,
  hidden_channels=32,
  out_channels=10,
  num_layers=16,
  scales=3,
  dropout=0.2,
  conv_type='gcnconv',
)

Code for additional message-passing protocols, as well as ready-to-use code for all the experiments from the paper, will be released in the near future.

City-networks

For information about the dataset see the original repo.

The training script is provided in main_city_networks.py. The command used to run the experiments:

model_layers=16
scales=4
dataset_name=paris

for seed in {0..4}; do
    python main.py --dataset $dataset_name --seed $seed --model imgcn --model-layers $model_layers --scales $scales --wandb --run-name $dataset_name-imgcn-$model_layers-layers-$scales-scales-$seed-seed
done

License

This project is released under the MIT license. Please see the LICENSE file for more information.

Citation

If you find this repository helpful, please consider citing:

@inproceedings{finder2025improving,
  title     = {Improving the Effective Receptive Field of Message-Passing Neural Networks},
  author    = {Finder, Shahaf E and Shapira Weber, Ron and Eliasof, Moshe and Freifeld, Oren and Treister, Eran},
  booktitle = {International Conference on Machine Learning},
  year      = {2025},
}