Graph and Geometric Deep Learning

Graph and geometric deep learning study neural networks for data beyond regular grids: graphs, molecules, meshes, manifolds, groups, and symmetries.

Paper Database

Year	Paper	Topic	Note
2016	Semi-Supervised Classification with Graph Convolutional Networks	GCN	Canonical graph convolution baseline.
2017	Inductive Representation Learning on Large Graphs	GraphSAGE	Sampling and aggregation for unseen nodes.
2017	Neural Message Passing for Quantum Chemistry	MPNN	General message-passing abstraction.
2017	Graph Attention Networks	GAT	Attention-weighted neighbor aggregation.
2018	How Powerful are Graph Neural Networks?	GIN / expressivity	Links GNN expressivity to graph isomorphism tests.
2020	SE(3)-Transformers	Equivariant attention	Roto-translation equivariance for 3D data.
2021	E(n) Equivariant Graph Neural Networks	EGNN	Efficient equivariance for geometric graphs.
2021	Geometric Deep Learning	Survey	Unifying view across grids, groups, graphs, and manifolds.

Step	Read
1	GCN and GraphSAGE.
2	MPNN and GAT.
3	GIN for expressivity.
4	SE(3)-Transformer and EGNN for equivariant geometry.
5	Geometric Deep Learning for the unifying map.