GNN (Graph Networks)

A GNN learns from things and the connections between them.

What it does

A GNN learns from things and the connections between them.

Node: A data point • Edge: A connection between two points

How it works

• Treat each thing as a node.
• Treat each connection as an edge.
• Let connected nodes share information.

Uses and applications

• A social app can suggest friends of friends.
• A delivery company can study roads and routes.
• A fraud system can find suspicious connections between accounts.

In short

Graph networks work with nodes and edges.

Related methods

• Neural Network: The layer-based learning that a GNN applies to graphs.
• Collaborative Filtering: Recommendation data is often modelled as a user-item graph.

Back to Neural networks and deep learning · All topics

The interactive demonstration and the Python walkthrough for this topic need JavaScript.