Glossary · Generative AI and LLMs
What is ScaNN? Definition and meaning
A library that uses anisotropic quantization to give accurate maximum-inner-product search.
Definition
ScaNN: A library that uses anisotropic quantization to give accurate maximum-inner-product search.
In short: Google's approximate nearest-neighbour library.
When it is used
High-throughput embedding retrieval.
Where it fits
ScaNN belongs to Vector indexes in the Generative AI and LLMs topic. These are the other terms in the same group:
- IVF-PQ: Partitions vectors into clusters and compresses them with product quantization to save memory.
- DiskANN: A graph-based index designed to serve very large vector collections from disk with low memory.
Learn more
The full guide to ScaNN explains how it works, when to use it and how it compares with the alternatives, with examples. The glossary lists every term on the site from A to Z.
Previous term: Sandboxing · Next term: Schema evolution