Master Graph Neural Networks from fundamentals to production. Learn DeepWalk, GCN, GAT, GraphSAGE, Graph Transformers, and more with hands-on Python implementations. New chapters on graph databases, LLMs, and foundation models.
Key Features
- Understand and implement core GNN architectures including GCN, GAT, GraphSAGE, and Graph Transformers
- Apply GNNs to real-world tasks like traffic forecasting, anomaly detection, recommender systems, and GraphRAG
- New coverage of graph databases, Graph Transformers, LLM-GNN integration, and Graph Foundation Models
Book Description
Graph Neural Networks have become essential tools for learning from relational and structured data. This second edition provides a comprehensive, hands-on guide implementing GNNs using Python and PyTorch Geometric. The book begins with graph theory fundamentals and data manipulation using NetworkX and PyTorch Geometric. You then explore shallow embedding methods—DeepWalk and Node2Vec—before progressing to core GNN architectures: Graph Convolutional Networks, Graph Attention Networks, and GraphSAGE. This edition introduces several new chapters reflecting the latest advances: Graph Transformers, integration of graph databases with GNNs, the convergence of LLMs and GNNs through GraphRAG, and the emerging paradigm of Graph Foundation Models. Existing chapters on expressiveness, link prediction, heterogeneous graphs, temporal GNNs, and explainability have been updated with current best practices. The final part puts theory into practice with end-to-end projects: traffic forecasting with A3T-GCN, anomaly detection with heterogeneous GNNs, and building a recommender system with LightGCN. By the end of this book, you will have the knowledge and practical skills to apply GNNs to your own graph-structured data problems.
What you will learn
- Master graph theory and manipulate graph data with NetworkX and PyG
- Create node embeddings using DeepWalk and Node2Vec
- Implement GCN, GAT, and GraphSAGE architectures from scratch
- Build and train Graph Transformers on molecular datasets
- Connect graph databases to PyG for scalable GNN pipelines
- Apply GNNs to traffic forecasting, anomaly detection, and recommendations
- Integrate GNNs with LLMs for graph-based retrieval-augmented generation
- Understand the emerging landscape of Graph Foundation Models
Who this book is for
This book is for machine learning practitioners, data scientists, and researchers who want to learn how to apply deep learning to graph-structured data. Software engineers working with network data, knowledge graphs, or recommendation systems will also benefit. A working knowledge of Python and familiarity with basic machine learning concepts (neural networks, training loops, loss functions) is expected. Prior experience with PyTorch is helpful but not required, as key concepts are introduced progressively.
