Neural diffusion PDEs, differential geometry, and graph neural networks

Name: Neural diffusion PDEs, differential geometry, and graph neural networks
Start: 2022-02-02T14:00:00-05:00
End: 2022-02-02T15:00:00-05:00
Location: Virtually

CMSA EVENTS: CMSA NEW TECHNOLOGIES IN MATHEMATICS

When: February 2, 2022

2:00 pm - 3:00 pm

Where: Virtually

Speaker: Michael Bronstein - University of Oxford / Twitter

In this talk, I will make connections between Graph Neural Networks (GNNs) and non-Euclidean diffusion equations. I will show that drawing on methods from the domain of differential geometry, it is possible to provide a principled view on such GNN architectural choices as positional encoding and graph rewiring as well as explain and remedy the phenomena of oversquashing and bottlenecks.

https://harvard.zoom.us/j/99651364593?pwd=Q1R0RTMrZ2NZQjg1U1ZOaUYzSE02QT09

Password: cmsa