GNN
GNN
A neural network designed to process graph-structured data. GNNs perform optimizable transformations on graph attributes (e.g., nodes, edges, global context) while preserving graph symmetries such as permutation invariance. GNNs utilize a “graph-in, graph-out” architecture that takes an input graph with information and progressively transforms it into an output graph with the same connectivity as that of the input graph.
Source: NIST AI 100-2e2025 | Category: