AI Roadmaps
← All roadmaps
AdvancedRoadmap 18 · 11 steps · 3h 26m of video

Graph Neural Networks

Learn from data shaped like networks: node embeddings, message passing, GCN, GAT, GraphSAGE, link prediction, and real uses.

Before this: Deep Learning. Linear algebra helps.

  1. 14 min
    Step 1

    Graph neural networks from the ground up

    Why graphs need their own kind of network.

    Graph Neural Networks - a perspective from the ground up

    Alex Foo

  2. 16 min
    Step 2

    Node embeddings and node2vec

    Use examples/node2vec.py to turn random walks into node embeddings.

    Graph Embeddings (node2vec) explained - How nodes get mapped to vectors

    Philipp Brunenberg

    pyg-team/pytorch_geometric★ 24.1k
  3. 11 min
    Step 3

    Message passing

    The loop every GNN layer repeats: gather, combine, update.

    Simple Message Passing on Graphs

    WelcomeAIOverlords

  4. 9 min
    Step 4

    Graph convolutional networks

    The simplest popular GNN layer, with the intuition.

    Graph Convolutional Networks (GCNs) made simple

    WelcomeAIOverlords

  5. 15 min
    Step 5

    Graph attention networks

    Use examples/gat.py to train a network that weights its neighbors.

    Understanding Graph Attention Networks

    DeepFindr

    pyg-team/pytorch_geometric★ 24.1k
  6. 13 min
    Step 6

    GraphSAGE

    Sample neighbors to train on graphs too big for memory.

    GraphSAGE: Inductive Representation Learning on Large Graphs (Graph ML Research Paper Walkthrough)

    TechViz - The Data Science Guy

  7. 3 min
    Step 7

    What GNNs can and cannot tell apart

    The Weisfeiler-Lehman test as a limit on expressiveness.

    Weisfeiler-Lehman WL Test for Graph Isomorphism explained visually & Message Passing NNs 2022

    Discover AI

  8. 1h 14m
    Step 8

    Hands-on with PyTorch Geometric

    Build and train a graph neural network with PyTorch Geometric.

    Graph Neural Networks (GNN) using Pytorch Geometric | Stanford University

    Lindsey AI

    pyg-team/pytorch_geometric★ 24.1k
  9. 17 min
    Step 9

    Link prediction

    Score candidate links using features computed from the graph.

    Stanford CS224W: ML with Graphs | 2021 | Lecture 2.2 - Traditional Feature-based Methods: Link

    Stanford Online

  10. 13 min
    Step 10

    Recommenders with GNNs

    Treat users and items as a bipartite graph.

    Recommender Systems using Graph Neural Networks

    DeepFindr

  11. 19 min
    Step 11

    GNNs that generate molecules

    Atoms as nodes, bonds as edges: a GNN proposes new molecules.

    Discovering New Molecules Using Graph Neural Networks by Rocío Mercado

    GAIA

Build it

Watching isn't learning. Pick one and put it on GitHub.

Where to go next