A neural network designed to learn from data shaped as a graph, using the links between nodes as part of what it learns.
Graph Neural Networks (GNNs) operate directly on graph structures such as Knowledge Graphs. Unlike LLMs, which process sequential text, GNNs learn from the connections between entities. That makes them useful for drug discovery, fraud detection, and logistics network optimization.
See also Large Language Model, Fine-Tuning and Model Inference.