/benchmarking-gnns

Repository for benchmarking graph neural networks

Primary LanguageJupyter NotebookMIT LicenseMIT

Benchmarking Graph Neural Networks


Updates

Nov 2, 2020

  • Project based on DGL 0.4.2. See the relevant dependencies defined in the environment yml files (CPU, GPU).
    • Numerical experiments report faster training times with DGL 0.4.2 compared to DGL 0.5.2.
    • For the version of the project compatible with DGL 0.5.2 and relevant dependencies, please use this branch.
  • Added ZINC-full dataset (249K molecular graphs) with scripts.

Jun 11, 2020

  • Second release of the project. Major updates :
    • Added experimental pipeline for Weisfeiler-Lehman-GNNs operating on dense rank-2 tensors.
    • Added a leaderboard for all datasets.
    • Updated PATTERN dataset.
    • Fixed bug for PATTERN and CLUSTER accuracy.
    • Moved first release to this branch.
  • New ArXiv's version of the paper.

Mar 3, 2020

  • First release of the project.

1. Benchmark installation

Follow these instructions to install the benchmark and setup the environment.


2. Download datasets

Proceed as follows to download the benchmark datasets.


3. Reproducibility

Use this page to run the codes and reproduce the published results.


4. Adding a new dataset

Instructions to add a dataset to the benchmark.


5. Adding a Message-passing GCN

Step-by-step directions to add a MP-GCN to the benchmark.


6. Adding a Weisfeiler-Lehman GNN

Step-by-step directions to add a WL-GNN to the benchmark.


7. Leaderboards

Leaderboards of GNN models on each dataset. Instructions to contribute to leaderboards.


8. Reference

ArXiv's paper

@article{dwivedi2020benchmarkgnns,
  title={Benchmarking Graph Neural Networks},
  author={Dwivedi, Vijay Prakash and Joshi, Chaitanya K and Laurent, Thomas and Bengio, Yoshua and Bresson, Xavier},
  journal={arXiv preprint arXiv:2003.00982},
  year={2020}
}