Timezone: »

 
Poster
Beltrami Flow and Neural Diffusion on Graphs
Benjamin Chamberlain · James Rowbottom · Davide Eynard · Francesco Di Giovanni · Xiaowen Dong · Michael Bronstein

Thu Dec 09 08:30 AM -- 10:00 AM (PST) @
in Poster Session 6 »

We propose a novel class of graph neural networks based on the discretized Beltrami flow, a non-Euclidean diffusion PDE. In our model, node features are supplemented with positional encodings derived from the graph topology and jointly evolved by the Beltrami flow, producing simultaneously continuous feature learning, topology evolution. The resulting model generalizes many popular graph neural networks and achieves state-of-the-art results on several benchmarks.

Keywords: Deep Learning · Graph Learning

Author Information

Benjamin Chamberlain (Twitter)
James Rowbottom (Twitter)
Davide Eynard (Università della Svizzera Italiana)
Francesco Di Giovanni (Twitter)
Xiaowen Dong (University of Oxford)
Michael Bronstein (Imperial College London / Twitter)

More from the Same Authors