Graph Neural Networks
Learning on graphs — representing nodes, edges, and whole graphs, the message-passing framework, graph convolutions, attention, and pooling for molecules, social networks, and knowledge graphs.
4 lessons
01Graphs as Data & Message Passing
When your data is a network — representing graphs for learning, why grids and sequences are special cases, and the message-passing framework that underlies every graph neural network.
Graph Convolutions & GraphSAGE
Turning message passing into a trainable layer — the Graph Convolutional Network, why the adjacency is normalized, and how GraphSAGE samples neighbors to scale and generalize to unseen nodes.
Graph Attention, Pooling & Over-smoothing
Learning which neighbors matter with graph attention, pooling node embeddings into a graph-level prediction, and why deep GNNs collapse — the over-smoothing problem and its fixes.
Graph Neural Networks Quiz
Check your understanding of graphs as data, message passing, graph convolutions, GraphSAGE, attention, pooling, and over-smoothing.