Tauseef FayyazA to Z
Glossary
Every concept in this sheet, plus the smaller terms it uses. Select a concept to open its card.
A
- Approximate nearest neighbours
- Search with an index that visits only a small part of the data and returns most, but not always all, of the true nearest neighbours.
C
- Choosing an embedding model
- The model that maps input to vectors.
- Cosine similarity
- The cosine of the angle between two vectors, from -1 (opposite) to 1 (same direction).
- Curse of dimensionality
- In high dimensions, distances between random points bunch together and classic space-partitioning structures such as k-d trees stop pruning much.
D
- Dimensions
- The number of values in each vector, fixed by the model, commonly between 384 and 3072.
- Dot product
- The sum of the element-wise products of two vectors.
E
- Embedding
- A fixed-length list of floating-point numbers that a model produces for a piece of text, an image or other data, placed so that similar meanings land close together in the vector space.
- Euclidean distance
- The straight-line distance between two points: the square root of the summed squared differences.
- Exact k-nearest neighbours
- Compare the query with every stored vector and keep the k closest.
N
- Normalisation
- Scaling a vector to length 1 by dividing it by its norm.
R
- Recall of an ANN index
- The share of the true top k neighbours, found by exact search, that the ANN search also returned. 0.95 means 95 of every 100 true neighbours were found.
S
- Semantic vs lexical search
- Semantic search compares meaning through embeddings.
T
- Top-k and similarity thresholds
- k is how many neighbours a query returns.
W
- When exact search is enough
- Below roughly tens of thousands of vectors, or after a filter narrows the candidates to a few thousand, a full scan is often fast enough.