Distance metrics
Available distance metricsโ
If not specified explicitly, the default distance metric in Weaviate is
cosine
. It can be set in the vectorIndexConfig field as part of the schema (here's an example adding a class to the schema) to any of the following types:
In all cases, larger distance values indicate lower similarity. Conversely, smaller distance values indicate higher similarity.
Name | Description | Definition | Range | Examples |
---|---|---|---|---|
cosine | Cosine (angular) distance. [See note 1 below] | 1 - cosine_sim(a,b) | 0 <= d <= 2 | 0 : identical vectors2 : Opposing vectors. |
dot | A dot product-based indication of distance. More precisely, the negative dot product. [See note 2 below] | -dot(a,b) | -โ < d < โ | -3 : more similar than -2 2 : more similar than 5 |
l2-squared | The squared euclidean distance between two vectors. | sum((a_i - b_i)^2) | 0 <= d < โ | 0 : identical vectors |
hamming | Number of differences between vectors at each dimensions. | sum(|a_i != b_i|) | 0 <= d < โ | 0 : identical vectors |
manhattan | The distance between two vector dimensions measured along axes at right angles. | sum(|a_i - b_i|) | 0 <= d < dims | 0 : identical vectors |
If you're missing your favorite distance type and would like to contribute it to Weaviate, we'd be happy to review your PR.
- If
cosine
is chosen, all vectors are normalized to length 1 at import/read time and dot product is used to calculate the distance for computational efficiency. - Dot Product on its own is a similarity metric, not a distance metric. As a result, Weaviate returns the negative dot product to stick with the intuition that a smaller value of a distance indicates a more similar result and a higher distance value indicates a less similar result.
Distance implementations and optimizationsโ
On a typical Weaviate use case the largest portion of CPU time is spent calculating vector distances. Even with an approximate nearest neighbor index - which leads to far fewer calculations - the efficiency of distance calculations has a major impact on overall performance.
You can use the following overview to find the best possible combination of distance metric and CPU architecture / instruction set.
Distance | linux/amd64 AVX2 | darwin/amd64 AVX2 | linux/amd64 AVX512 | linux/arm64 | darwin/arm64 |
---|---|---|---|---|---|
cosine | optimized | optimized | no SIMD | no SIMD | no SIMD |
dot | optimized | optimized | no SIMD | no SIMD | no SIMD |
l2-squared | optimized | optimized | no SIMD | no SIMD | no SIMD |
hamming | no SIMD | no SIMD | no SIMD | no SIMD | no SIMD |
manhattan | no SIMD | no SIMD | no SIMD | no SIMD | no SIMD |
If you like dealing with Assembly programming, SIMD, and vector instruction sets we would love to receive your contribution for one of the combinations that have not yet received an SIMD-specific optimization.
Distance fields in the APIsโ
The distance
is exposed in the APIs in two ways:
- Whenever a vector search is involved, the distance can be displayed as part of the results, for example using
_additional { distance }
- Whenever a vector search is involved, the distance can be specified as a limiting criterion, for example using
nearVector({distance: 1.5, vector: ... })
Note: The distance
field was introduced in v1.14.0
. In previous versions, only certainty
(see below) was available.
Distance vs Certaintyโ
Prior to version v1.14
only certainty
was available in the APIs. The
original ideas behind certainty was to normalize the distance score into a
value between 0 <= certainty <= 1
, where 1 would represent identical vectors
and 0 would represent opposite vectors.
This concept is however unique to cosine
distance. With other distance
metrics, scores may be unbounded. As a result the preferred way is to use
distance
in favor of certainty
.
For backward compatibility, certainty
can still be used when the distance is
cosine
. If any other distance is selected certainty
cannot be used.
See also distance and certainty _additional{} properties.
More resourcesโ
If you can't find the answer to your question here, please look at the:
- Frequently Asked Questions. Or,
- Knowledge base of old issues. Or,
- For questions: Stackoverflow. Or,
- For more involved discussion: Weaviate Community Forum. Or,
- We also have a Slack channel.