Graph Anisotropic Diffusion for Molecules

Abstract

Traditional Graph Neural Networks (GNNs) rely on message passing, which amounts to permutation-invariant local aggregation of neighbour features. Such a process is isotropic and there is no notion of ‘direction’ on the graph. We present a new GNN architecture called Graph Anisotropic Diffusion. Our model alternates between linear diffusion, for which a closed-form solution is available, and local anisotropic filters to obtain efficient multi-hop anisotropic kernels. We test our model on two common molecular property prediction benchmarks (ZINC and QM9) and show its competitive performance.

Cite

Text

Elhag et al. "Graph Anisotropic Diffusion for Molecules." ICLR 2022 Workshops: MLDD, 2022.

Markdown

[Elhag et al. "Graph Anisotropic Diffusion for Molecules." ICLR 2022 Workshops: MLDD, 2022.](https://mlanthology.org/iclrw/2022/elhag2022iclrw-graph-a/)

BibTeX

@inproceedings{elhag2022iclrw-graph-a,
  title     = {{Graph Anisotropic Diffusion for Molecules}},
  author    = {Elhag, Ahmed A. A. and Corso, Gabriele and Stärk, Hannes and Bronstein, Michael M.},
  booktitle = {ICLR 2022 Workshops: MLDD},
  year      = {2022},
  url       = {https://mlanthology.org/iclrw/2022/elhag2022iclrw-graph-a/}
}