Graph neural network
A graph neural network is a neural-network model that learns representations or predictions from graph-structured data by combining information from connected entities.
Graph Neural Networks in Action
A graph neural network architecture specifies the layers, computations, and connectivity used by a graph neural network.
- How it is used
- Models relationship-centric data for node, edge, or graph predictions when connections contribute useful information.
- How deep it goes
- Builds from graph representation to message-passing layers and a coded two-layer GCN; defers detailed convolution treatment.
- Examples
- Uses a two-layer GCN embedding model, sparse user-item interactions, and atom-bond graphs for molecular prediction.
- What you can do after
- Build it from scratch as a basic two-layer GCN in PyTorch Geometric and evaluate whether graph relationships justify its use.
Read how Graph Neural Networks in Action explains graph neural network →
Financial AI in Practice
A neural network that learns from graph-structured data by aggregating information across connected nodes.
- How it is used
- Scores connected financial activity for fraud and supplies relationship-based signals for investment recommendations.
- How deep it goes
- Works through graph preparation, a two-layer PyTorch model, training, and ranked output, while explicitly stopping short of production-grade design.
- Examples
- Uses transactions linked by shared users or devices, a Normal/Fraud classifier, and a shared-holding recommendation scenario.
- What you can do after
- Build it from scratch as a simplified PyTorch transaction GNN and inspect its highest-ranked fraud probabilities.
Read how Financial AI in Practice explains graph neural network →
Knowledge Graphs and LLMs in Action
A graph neural network is a neural network architecture that learns representations by propagating information among nodes and edges of a graph.
- How it is used
- Deploys graph structure and node attributes to learn representations for classification, link prediction, and other graph-based machine-learning tasks.
- How deep it goes
- Explains message passing, trainable update rules, embeddings, and encoder–decoder workflows, without deriving advanced architectures.
- Examples
- Uses financial transaction graphs for AML classification and user–movie interaction graphs for recommendation link prediction.
- What you can do after
- Apply it via PyTorch Geometric to build and evaluate node-classification or link-prediction systems.
Read how Knowledge Graphs and LLMs in Action explains graph neural network →
Optimization Algorithms: AI techniques for design, planning, and control problems
A graph neural network is a neural network architecture that learns representations and predictions from graph-structured data.
- How it is used
- Applies graph representations and predictions to combinatorial optimization problems such as TSP, VRP, SAT, MaxCut, and MIS.
- How deep it goes
- Explains encoding, message passing, prediction, search, and learning objectives, while deferring training implementation details to substantial code and preprocessing.
- Examples
- Uses TSP, pretrained models, node and edge probabilities, Concorde solutions, beam search, and a Gurobi-assisted implementation.
- What you can do after
- Apply it via pretrained models to predict and search feasible TSP solutions.
Machine Learning Algorithms in Depth
A graph neural network is a neural-network architecture that computes representations by propagating information over graph-structured data.
- How it is used
- Uses graph-structured relationships and node features to classify documents in a citation network.
- How deep it goes
- Defines adjacency and feature matrices, distinguishes classification targets, and provides a GCN implementation listing.
- Examples
- Uses CORA documents, citation edges, word-indicator features, and a t-SNE view of a hidden representation.
- What you can do after
- Apply it via Spektral’s GCNConv with TensorFlow/Keras to classify nodes in a citation graph.
Read how Machine Learning Algorithms in Depth explains graph neural network →
Related concepts
- graph attention networkincludes · A graph attention network is a graph neural network that uses attention weights to aggregate information from neighboring nodes, allowing their contributions to differ.
- graphuses · A graph is a mathematical structure consisting of vertices connected by edges that represent relationships or connections.
- PyTorch Geometricused by · PyTorch Geometric is an open-source Python library that extends PyTorch with data structures and operators for graph and geometric machine learning.
- node classificationuses · Node classification is a graph-learning task that predicts one or more discrete category labels for individual nodes in a graph.
- link predictionuses · Link prediction is a graph-learning task that estimates whether a connection exists or may occur between nodes, often by scoring candidate links.
- graph convolutional networkincludes · A graph convolutional network is a graph neural network that computes representations by transforming and aggregating information from graph neighborhoods.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.