Graph convolution
Graph convolution is a graph-neural-network operation that updates node representations by transforming node features and aggregating information from neighboring nodes.
Graph Neural Networks in Action
Graph convolution combines a node's representation with information aggregated from its graph neighborhood.
- How it is used
- Deploys graph convolution inside GNN architectures to learn node representations and support graph-based classification.
- How deep it goes
- Explains neighborhood aggregation, stacked layers, activations, and outputs, while stopping short of a detailed spectral derivation.
- Examples
- Uses two-layer GCN and GraphSAGE classes with edge_index, ReLU, embedding return, and log-softmax classification output.
- What you can do after
- Build it from scratch with PyTorch Geometric layers, stacked activations, embedding outputs, and classification outputs.
Read how Graph Neural Networks in Action explains graph convolution →
Financial AI in Practice
A neural-network layer that transforms node features and combines information from neighboring nodes in a graph.
- How it is used
- Uses neighborhood information to learn relational features for financial fraud detection.
- How deep it goes
- Explains linear transformation, adjacency-based aggregation, and feature combination, while deferring specialized layers for large, dynamic graphs.
- Examples
- Demonstrates transaction features based on amount and time, with graph edges for shared users or devices.
- What you can do after
- Build from scratch a simplified two-layer graph neural network for transaction fraud classification.
Read how Financial AI in Practice explains graph convolution →
Related concepts
- GCNConvimplemented by · GCNConv is a PyTorch Geometric graph-convolution layer that updates node representations through degree-normalized aggregation of neighboring node features.
- embeddingproduces · A numerical vector representation of an item or input that encodes features or relationships useful for computational tasks.
- data aggregationuses · Data aggregation is the combination of multiple values, records, observations, or results into a consolidated or summarized output.
- linear layerhas part · A linear layer is a neural-network layer that maps input features to outputs by a learned affine transformation, typically connecting each output to every input feature.
- subgraphused by · A subgraph is a graph whose vertices and edges are subsets of those of another graph, with every included edge connecting included vertices.
- graph convolutional networkused by · 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.