Graph Neural Network
Also called GNN
A network built for data shaped like a network — nodes and connections — that learns by passing messages between neighbors.
Think of it like
Like gossip spreading through a friend group: each person updates their view based on what their direct friends say.
Example
Fraud detection systems use GNNs over transaction graphs to spot rings of accounts that look normal alone but suspicious together.
How it actually works
Where transformers assume a sequence and CNNs assume a grid, GNNs assume an arbitrary graph, and each node updates itself by aggregating messages from its neighbors over several rounds. That fits molecules, social networks, and recommendation graphs naturally. They’re powerful on relational data but can struggle to pass information across very distant nodes.
For product teams
The right tool when your data is fundamentally about connections, not sequences.
For engineers
A network operating on graph-structured data via iterative neighborhood message passing.
Related
- Transformer — The sequence-oriented alternative.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome