Categories:

Vector functions

VECTOR_NORMALIZE

Normalizes a VECTOR in the L2 vector space, giving its elements values in the range of [0,1] and giving it a magnitude of 1.

See also:

Vector Embeddings, VECTOR_TRUNCATE

Syntax

VECTOR_NORMALIZE( <vector> )
Copy

Arguments

vector

A single VECTOR value to normalize.

Returns

Returns a VECTOR normalized to the L2 space, with values of type FLOAT.

Usage notes

  • Returns NULL when the input is NULL.

  • Vector functions are optimized in a way that can reduce floating point precision. This function’s results have a margin of error up to 1e-4.

Examples

This example demonstrates normalizing the vector [1, 2, 3]:

SELECT VECTOR_NORMALIZE([1, 2, 3]::VECTOR(INT, 3));
Copy
[0.267261, 0.534522, 0.801784]

This example shows how to re-normalize a truncated vector. The original vector is produced by EMBED_TEXT_768 with the snowflake-arctic-embed-m-v1.5 model, and then truncated to 256 elements. The truncated vector is then normalized:

VECTOR_NORMALIZE(
    VECTOR_TRUNCATE(
        SNOWFLAKE.CORTEX.EMBED_TEXT_768(
            'snowflake-arctic-embed-m-v1.5',
            'Analytical databases are typically column-oriented rather than row-oriented'
        ),
        256
    )
);
Copy
Language: English