Dive into the world of Graph Convolutional Networks (GCNs) in this comprehensive 58-minute lecture. Explore the concept of tensors and representations living on vertices and edges, and understand the connection between GCNs and Convolutional Neural Networks (CNNs) on grids. Learn about residual gated GCNs and gain insights into domain sparsity. Discover practical implementation techniques using PyTorch and the Deep Graph Library (DGL). Benefit from self-learning resources provided by experts Xavier Bresson and Jure Leskovec, and build upon knowledge from previous lectures on Graph Transformer Networks (GTNs).
Overview
Syllabus
– Welcome to class
– Recap from lecture 10 → Graph Transformer Networks GTNs
– Today plan: tensors/representations living on vertices and edges
– Self-learning resources with Xavier Bresson and Jure Leskovec
– Graph Convolutional Networks GCNs
– Connection with Convolutional Nets CNNs on grids
– Residual gated GCNs
– Domain sparsity note
– PyTorch implementation using Deep Graph Library DGL
– And that was it!
Taught by
Alfredo Canziani
