MetricType

The metric type used to measure the distance/similarity between vectors.

public enum MetricType
{
    Invalid = 0,
    L2 = 1,
    Ip = 2,
    Cosine = 3,
    Hamming = 4,
    Jaccard = 5,
    MhJaccard = 6,
    Bm25 = 7,
    MaxSim = 8,
    MaxSimCosine = 9,
    MaxSimIp = 10,
    MaxSimL2 = 11,
    MaxSimJaccard = 12,
    MaxSimHamming = 13
}

Constants:

  • Invalid (0)

    An invalid metric type.

  • L2 (1)

    Squared Euclidean distance (smaller is more similar). For float vectors.

  • Ip (2)

    Inner product (larger is more similar). For float vectors.

  • Cosine (3)

    Cosine similarity (larger is more similar). For float vectors.

  • Hamming (4)

    Hamming distance (smaller is more similar). For binary vectors.

  • Jaccard (5)

    Jaccard distance (smaller is more similar). For binary vectors.

  • MhJaccard (6)

    Modified Jaccard distance (smaller is more similar). For binary vectors.

  • Bm25 (7)

    The BM25 ranking metric for full-text search on sparse vectors. Only valid for sparse float vectors; larger BM25 scores are more similar.

  • MaxSim (8)

    Max-sim cosine similarity (larger is more similar). Equal to MaxSimCosine. Applies to float vectors stored inside a struct field.

  • MaxSimCosine (9)

    Max-sim cosine similarity (larger is more similar). Applies to float vectors inside a struct field.

  • MaxSimIp (10)

    Max-sim inner product. Applies to float vectors inside a struct field.

  • MaxSimL2 (11)

    Max-sim squared Euclidean distance. Applies to float vectors inside a struct field.

  • MaxSimJaccard (12)

    Max-sim Jaccard similarity (larger is more similar). Applies to binary vectors inside a struct field.

  • MaxSimHamming (13)

    Max-sim Hamming similarity (smaller is more similar). Applies to binary vectors inside a struct field.

Example

using Milvus.Client.V2;
using Milvus.Client.V2.Requests.Dql;
using Milvus.Client.V2.Responses.Dql;
using Milvus.Client.V2.Types;

using var client = new MilvusClientV2(new ConnectConfig
{
    Uri = "localhost:19530",
    Username = "root",
    Password = "Milvus"
});

SearchResp results = await client.SearchAsync(new SearchReq
{
    CollectionName = "my_collection",
    VectorFieldName = "embedding",
    Vectors = [new[] { 0.1f, 0.2f, 0.3f, 0.4f }],
    MetricType = MetricType.Cosine,
    Limit = 5
});

Try Managed Milvus for Free

Zilliz Cloud is hassle-free, powered by Milvus and 10x faster.

Get Started
Feedback

Was this page helpful?