Graph Anisotropic Diffusion
This work addresses the limitation of lacking directionality in GNNs for tasks like molecular property prediction, though it appears incremental as it builds on existing GNN frameworks.
The paper tackled the problem of isotropic message passing in Graph Neural Networks (GNNs) by introducing Graph Anisotropic Diffusion, a new architecture that incorporates directionality on graphs, and demonstrated competitive performance on molecular property prediction benchmarks ZINC and QM9.
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.