HNSW
Graph index that finds approximate nearest neighbours in logarithmic time.
Overview
HNSW builds a layered graph where each vector links to a few close neighbours, and a search descends through the layers, narrowing in. It solves the problem that exact nearest-neighbour search is linear in the collection size, which stops being acceptable well before a corpus stops growing. It exists as the index almost every vector database uses by default, at the cost of recall that is high but not perfect.
Technology Graph
How HNSW connects to the rest of the stack.
Used by
- Typesense database Search Engines HNSW used by Typesense High ✓ source
Applied to
- AI Search application Applications HNSW used for AI Search High ✓ source
- Embedding concept Techniques HNSW used for Embedding High ✓ source
Classification
- FAISS library Vector Databases HNSW part of FAISS High ✓ source
- Qdrant database Vector Databases HNSW part of Qdrant High ✓ source