Implementation of Directional Graph Networks in PyTorch and DGL.
We provide the implementation of the Directional Graph Networks (DGN) in PyTorch and DGL frameworks, along with scripts for running real-world benchmarks. The repository is organised as follows:
modelscontains:pytorchcontains the various GNN models implemented in PyTorch: the implementation of the aggregators, the scalers, the DGN layer and the directional aggregation matrix (eigen_agg).dglcontains the DGN model implemented via the DGL library: aggregators, scalers, and DGN layer.layers.pycontains general NN layers used by the various models
realworld_benchmarkcontains various scripts from Benchmarking GNNs and Open Graph Benchmark to download the real-world benchmarks and train the DGN on them. Inrealworld_benchmark/README.mdwe provide instructions for runnning the experiments.
@article{beaini2020directional,
title={Directional graph networks},
author={Beaini, Dominique and Passaro, Saro and L{\'e}tourneau, Vincent and Hamilton, William L and Corso, Gabriele and Li{\`o}, Pietro},
journal={arXiv preprint arXiv:2010.02863},
year={2020}
}
MIT