kNN Search

k-nearest-neighbour (kNN) search functionality is a way to find entries based on their distance to a specific vector.

Function syntax

The general function syntax for kNN is: distance(<target column>, <vector to compare with>, <metric> [, <weights>]).

KNN syntax

To express a k-Nearest-Neighbour search in SQL, you can combine the distance function with a LIMIT and ORDER BY clause.

Examples

SELECT id, distance(arraycolumn, ARRAY[...], 'L2') as dist FROM tablewitharraycolumn ORDER BY dist ASC LIMIT 100;
SELECT id, distance(arraycolumn, ARRAY[...], 'L2', ARRAY[...]) as dist FROM tablewitharraycolumn ORDER BY dist ASC LIMIT 100;

Supported metrics

Currently the following metrics are supported:

• L1
• L2
• L2 squared
• Cosine
• ChiSquared