neighbor

Definition

A data point close to another under the chosen metric. The idea spans classical k-nearest-neighbour classification, which does no training at all and simply consults similar examples, through to vector databases retrieving candidates for a query. Quality is decided by two choices behind it: the embedding, and whether distance is cosine, dot product, or Euclidean.

Explainers using this term

• Build Your Own Vector DB — From Brute Force to HNSWEfficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs
• Hashing and Nearest-Neighbor Search — The Groundwork Under Vector SearchEfficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs
• Graph Algorithms from Scratch — Shortest Paths and Where They LeadEfficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs