Chapters¶
This textbook covers 300 concepts across 27 chapters in 6 parts. Each chapter respects concept dependencies — every prerequisite is introduced before it is needed.
Chapter Overview¶
-
Chapter 0: Math and Programming Prerequisites (20 concepts)
Math, PyTorch, and graph library setup -
Chapter 1: Introduction to Graphs and Networks (33 concepts)
Introduction to graph structure and network science -
Chapter 2: Graph Properties and Traditional ML Features (18 concepts)
Classical graph features and kernels -
Chapter 3: Link Analysis and PageRank (7 concepts)
PageRank, random walks, and link analysis -
Chapter 4: Node Embeddings: DeepWalk and node2vec (17 concepts)
DeepWalk, node2vec, and shallow embeddings -
Chapter 5: Label Propagation and Semi-Supervised Learning (6 concepts)
Label propagation and semi-supervised learning -
Chapter 6: GNN Foundations: Message Passing and GCN (23 concepts)
GCN, message passing, spectral theory -
Chapter 7: GNN Design Space: GraphSAGE and GAT (22 concepts)
GraphSAGE, GAT, pooling, and task types -
Chapter 8: GNN Training, Augmentation, and Practical Tips (18 concepts)
Loss functions, normalization, and self-supervised GNN training -
Chapter 9: Theory of GNNs: Expressiveness and the WL Test (13 concepts)
WL test, expressiveness, over-smoothing, over-squashing -
Chapter 10: Designing Powerful Encoders: GIN and Beyond (9 concepts)
GIN, higher-order GNNs, PNA, stochastic depth -
Chapter 11: Graph Transformers (13 concepts)
Graphormer, GPS, SAN, and positional encodings -
Chapter 12: Knowledge Graph Embeddings (20 concepts)
TransE, RotatE, ComplEx, and relation pattern geometry -
Chapter 13: Reasoning over Knowledge Graphs (5 concepts)
Multi-hop query answering and neural path reasoning -
Chapter 14: Knowledge Graph Foundation Models (6 concepts)
ULTRA, InGram, and pre-training for KG transfer -
Chapter 15: Heterogeneous Graphs (8 concepts)
R-GCN, HGT, HAN, and meta-path reasoning -
Chapter 16: GNNs for Recommender Systems (6 concepts)
LightGCN, PinSage, and bipartite graph recommendation -
Chapter 17: Relational Deep Learning (5 concepts)
Relational databases as heterogeneous graphs -
Chapter 18: Community Structure in Networks (8 concepts)
Louvain, BigCLAM, and spectral community detection -
Chapter 19: Frequent Subgraph Mining (4 concepts)
SPMiner and neural subgraph pattern matching -
Chapter 20: Scaling GNNs to Billion-Node Graphs (9 concepts)
Neighbor sampling, Cluster-GCN, and SIGN -
Chapter 21: Deep Generative Models for Graphs (11 concepts)
GraphRNN, GCPN, VGAE, and molecular generation -
Chapter 22: Temporal and Dynamic Graphs (5 concepts)
TGN, TGAT, and learning on evolving graphs -
Chapter 23: LLMs and GNNs: Text-Attributed Graphs and Joint Training (4 concepts)
LLM+GNN integration, text-attributed graphs, OFA -
Chapter 24: Advanced GNN Topics: In-Context Learning and Uncertainty (7 concepts)
Self-supervised GNNs, DGI, contrastive learning, in-context learning -
Chapter 25: Agents and Graphs (3 concepts)
Agent memory graphs and tool-use workflows -
Chapter 26: Conclusion: The GNN Design Space and Open Problems (0 concepts)
GNN design space synthesis and open problems
How to Use This Textbook¶
Follow the chapters in order for the full learning path. Each chapter's Prerequisites section lists which earlier chapters must be completed first. The Learning Graph lets you visualize concept dependencies interactively if you want to plan a non-linear reading path.