#71 - ZAK JOST (Graph Neural Networks + Geometric DL) [UNPLUGGED]
Machine Learning Street Talk (MLST)
00:00
Advancing Graph Neural Networks
This chapter explores the interplay between cellular automata and graph neural networks, focusing on the transformation of discrete processes into continuous representations. It highlights innovative techniques such as label propagation and tree-based models, showcasing their effectiveness against complex GNNs. Additionally, the discussion examines the capabilities and limitations of GNNs, particularly regarding graph isomorphism detection through new approaches and aggregation functions.
Transcript
Play full episode
