In this post, we will describe how Vector databases work and also build a simple Vector Search library in under 200 lines of Rust code. All code can be found in this Github repo. The basic idea behind all vector search algorithms is the same – do some pre-processing to identify points that are close enough to each other.
Table of contents
Introduction to Vectors (aka Embeddings)How to Find Approximate Nearest Neighbours?Why is FANN so fast?More Real World ConsiderationsConclusionAbout Fennel2 Comments
Sort: