Graph Representation Learning
Petar discusses the challenges of reusing edge messages in graph representation learning, highlighting the limitations from a categorical perspective. By splitting the message function into two streams for edges and nodes, significant empirical benefits are observed in extrapolating edge-centric algorithms.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#85 Dr. Petar Veličković (Deepmind) - Categories, Graphs, Reasoning [NEURIPS22 UNPLUGGED]
Related Questions