Geometric Deep Learning: GNNs Beyond Permutation Equivariance

Casting graph neural networks (GNNs) within the Geometric Deep Learning blueprint, then demonstrating how we can use the blueprint to extend GNNs beyond the notion of permutation equivariance.
Guest Lecture at the Machine Learning with Graphs (CS224W) course, Stanford University, 30 November 2021

Slide deck: https://petar-v.com/talks/5G-CS224W.pdf