Timezone: »

 
Poster
Autobahn: Automorphism-based Graph Neural Nets
Erik Thiede · Wenda Zhou · Risi Kondor

Wed Dec 08 04:30 PM -- 06:00 PM (PST) @
in Poster Session 4 »

We introduce Automorphism-based graph neural networks (Autobahn), a new family of graph neural networks. In an Autobahn, we decompose the graph into a collection of subgraphs and apply local convolutions that are equivariant to each subgraph's automorphism group. Specific choices of local neighborhoods and subgraphs recover existing architectures such as message passing neural networks. Our formalism also encompasses novel architectures: as an example, we introduce a graph neural network that decomposes the graph into paths and cycles. The resulting convolutions reflect the natural way that parts of the graph can transform, preserving the intuitive meaning of convolution without sacrificing global permutation equivariance. We validate our approach by applying Autobahn to molecular graphs, where it achieves results competitive with state-of-the-art message passing algorithms.

Keywords: Deep Learning · Graph Learning

Author Information

Erik Thiede (Flatiron Institute)
Wenda Zhou (Columbia University)
Risi Kondor (Flatiron Institute)

Risi Kondor joined the Flatiron Institute in 2019 as a Senior Research Scientist with the Center for Computational Mathematics. Previously, Kondor was an Associate Professor in the Department of Computer Science, Statistics, and the Computational and Applied Mathematics Initiative at the University of Chicago. His research interests include computational harmonic analysis and machine learning. Kondor holds a Ph.D. in Computer Science from Columbia University, an MS in Knowledge Discovery and Data Mining from Carnegie Mellon University, and a BA in Mathematics from the University of Cambridge. He also holds a diploma in Computational Fluid Dynamics from the Von Karman Institute for Fluid Dynamics and a diploma in Physics from Eötvös Loránd University in Budapest.

More from the Same Authors