graph

Definition

Nodes connected by edges, and the methods that operate on them. The word points at two different things depending on context: the data itself — molecules, social networks, knowledge graphs — or the internal computation graph a framework builds to run backpropagation. Graphs matter because a lot of real data has no natural sequence or grid ordering, which is exactly what transformers and CNNs assume.

Explainers using this term

• Graph Neural Networks from Scratch — Learning from ConnectionsSemi-Supervised Classification with Graph Convolutional Networks
• GraphRAG from Scratch — Where Knowledge Graphs Meet RetrievalFrom Local to Global: A Graph RAG Approach to Query-Focused Summarization