Michael Bronstein: Physics-Inspired Graph Neural Networks

Explore physics-inspired "continuous" learning models for graph neural networks in this lecture by Michael Bronstein. Delve into the limitations of the message-passing paradigm and discover how tools from differential geometry, algebraic topology, and differential equations can revolutionize graph machine learning. Learn about the theoretical links between graph neural networks and the Weisfeiler-Lehman hierarchy, and understand why moving beyond the "node-and-edge"-centric approach may be crucial for future advancements in the field. Gain insights into the potential applications of these novel techniques, ranging from particle physics to protein design.