随机抽样 Compatible with Milvus 2.6.x

在处理大规模数据集时，您往往不需要处理所有数据来获得洞察力或测试过滤逻辑。随机抽样提供了一种解决方案，让您可以处理数据中具有统计代表性的子集，从而大大减少查询时间和资源消耗。

随机抽样在分段级别上操作，在确保高效性能的同时，还能在整个 Collection 的数据分布中保持样本的随机性。

主要用例

• 数据探索：以最少的资源使用量快速预览 Collection 的结构和内容

• 开发测试：在全面部署之前，在可管理的数据样本上测试复杂的过滤逻辑

• 资源优化：降低探索性查询和统计分析的计算成本

语法

Python

filter = "RANDOM_SAMPLE(sampling_factor)"

Java

String filter = "RANDOM_SAMPLE(sampling_factor)"

Go

filter := "RANDOM_SAMPLE(sampling_factor)"

NodeJS

// node

cURL

# restful

参数

• sampling_factor:取样系数，取样范围为 (0，1)，不包括边界。例如，RANDOM_SAMPLE(0.001) 会选择大约 0.1% 的结果。

重要规则

• 表达式不区分大小写 (RANDOM_SAMPLE 或random_sample)

• 取样因子必须在 (0, 1) 范围内，不包括边界

与其他筛选器结合使用

随机抽样操作符必须与其他过滤表达式相结合，使用逻辑AND 。组合过滤器时，Milvus 首先应用其他条件，然后对结果集执行随机抽样。

Python

# Correct: Filter first, then sample
filter = 'color == "red" AND RANDOM_SAMPLE(0.001)'
# Processing: Find all red items → Sample 0.1% of those red items

# Incorrect: OR doesn't make logical sense
filter = 'color == "red" OR RANDOM_SAMPLE(0.001)'  # ❌ Invalid logic
# This would mean: "Either red items OR sample everything" - which is meaningless

Java

// Correct: Filter first, then sample
String filter = 'color == "red" AND RANDOM_SAMPLE(0.001)';
// Processing: Find all red items → Sample 0.1% of those red items

// Incorrect: OR doesn't make logical sense
String filter = 'color == "red" OR RANDOM_SAMPLE(0.001)';  // ❌ Invalid logic
// This would mean: "Either red items OR sample everything" - which is meaningless

Go

// Correct: Filter first, then sample
filter := 'color == "red" AND RANDOM_SAMPLE(0.001)'
// Processing: Find all red items → Sample 0.1% of those red items

filter := 'color == "red" OR RANDOM_SAMPLE(0.001)' // ❌ Invalid logic
// This would mean: "Either red items OR sample everything" - which is meaningless

NodeJS

// node

cURL

# restful

示例

示例 1：数据探索

快速预览您的 Collection 结构：

Python

from pymilvus import MilvusClient

client = MilvusClient(uri="http://localhost:19530")

# Sample approximately 1% of the entire collection
result = client.query(
    collection_name="product_catalog",
    filter="RANDOM_SAMPLE(0.01)",
    output_fields=["id", "product_name"],
    limit=10
)

print(f"Sampled {len(result)} products from collection")

Java

import io.milvus.v2.client.*;
import io.milvus.v2.service.vector.request.QueryReq
import io.milvus.v2.service.vector.request.QueryResp

ConnectConfig config = ConnectConfig.builder()
        .uri("http://localhost:19530")
        .build();
MilvusClientV2 client = new MilvusClientV2(config);

QueryReq queryReq = QueryReq.builder()
        .collectionName("product_catalog")
        .filter("RANDOM_SAMPLE(0.01)")
        .outputFields(Arrays.asList("id", "product_name"))
        .limit(10)
        .build();

QueryResp queryResp = client.query(queryReq);

List<QueryResp.QueryResult> results = queryResp.getQueryResults();
for (QueryResp.QueryResult result : results) {
    System.out.println(result.getEntity());
}

Go

import (
    "context"
    "fmt"

    "github.com/milvus-io/milvus/client/v2/entity"
    "github.com/milvus-io/milvus/client/v2/milvusclient"
)

ctx, cancel := context.WithCancel(context.Background())
defer cancel()

milvusAddr := "localhost:19530"
client, err := milvusclient.New(ctx, &milvusclient.ClientConfig{
    Address: milvusAddr,
})
if err != nil {
    fmt.Println(err.Error())
    // handle error
}
defer client.Close(ctx)

resultSet, err := client.Query(ctx, milvusclient.NewQueryOption("product_catalog").
    WithFilter("RANDOM_SAMPLE(0.01)").
    WithOutputFields("id", "product_name"))
if err != nil {
    fmt.Println(err.Error())
    // handle error
}

fmt.Println("id: ", resultSet.GetColumn("id").FieldData().GetScalars())
fmt.Println("product_name: ", resultSet.GetColumn("product_name").FieldData().GetScalars())

NodeJS

// node

cURL

# restful

示例 2：过滤与随机抽样相结合

在可管理的子集上测试过滤逻辑：

Python

# First filter by category and price, then sample 0.5% of results
filter_expression = 'category == "electronics" AND price > 100 AND RANDOM_SAMPLE(0.005)'

result = client.query(
    collection_name="product_catalog",
    filter=filter_expression,
    output_fields=["product_name", "price", "rating"],
    limit=10
)

print(f"Found {len(result)} electronics products in sample")

Java

String filter = "category == \"electronics\" AND price > 100 AND RANDOM_SAMPLE(0.005)";

QueryReq queryReq = QueryReq.builder()
        .collectionName("product_catalog")
        .filter(filter)
        .outputFields(Arrays.asList("product_name", "price", "rating"))
        .limit(10)
        .build();

QueryResp queryResp = client.query(queryReq);

Go

filter := "category == \"electronics\" AND price > 100 AND RANDOM_SAMPLE(0.005)"

resultSet, err := client.Query(ctx, milvusclient.NewQueryOption("product_catalog").
    WithFilter(filter).
    WithOutputFields("product_name", "price", "rating"))
if err != nil {
    fmt.Println(err.Error())
    // handle error
}

NodeJS

// node

cURL

# restful

示例 3：快速分析

对过滤后的数据进行快速统计分析：

Python

# Get insights from ~0.1% of premium customer data
filter_expression = 'customer_tier == "premium" AND region == 'North America' AND RANDOM_SAMPLE(0.001)'

result = client.query(
    collection_name="customer_profiles",
    filter=filter_expression,
    output_fields=["purchase_amount", "satisfaction_score", "last_purchase_date"],
    limit=10
)

# Analyze sample for quick insights
if result:
    average_purchase = sum(r["purchase_amount"] for r in result) / len(result)
    average_satisfaction = sum(r["satisfaction_score"] for r in result) / len(result)

    print(f"Sample size: {len(result)}")
    print(f"Average purchase amount: ${average_purchase:.2f}")
    print(f"Average satisfaction score: {average_satisfaction:.2f}")

Java

String filter = "customer_tier == \"premium\" AND region == \"North America\" AND RANDOM_SAMPLE(0.001)";

QueryReq queryReq = QueryReq.builder()
        .collectionName("customer_profiles")
        .filter(filter)
        .outputFields(Arrays.asList("purchase_amount", "satisfaction_score", "last_purchase_date"))
        .limit(10)
        .build();

QueryResp queryResp = client.query(queryReq);

Go

filter := "customer_tier == \"premium\" AND region == \"North America\" AND RANDOM_SAMPLE(0.001)"

resultSet, err := client.Query(ctx, milvusclient.NewQueryOption("customer_profiles").
    WithFilter(filter).
    WithOutputFields("purchase_amount", "satisfaction_score", "last_purchase_date"))
if err != nil {
    fmt.Println(err.Error())
    // handle error
}

NodeJS

// node

cURL

# restful

示例 4：与向量搜索相结合

在过滤搜索场景中使用随机抽样：

Python

# Search for similar products within a sampled subset
search_results = client.search(
    collection_name="product_catalog",
    data=[[0.1, 0.2, 0.3, 0.4, 0.5]],  # query vector
    filter='category == "books" AND RANDOM_SAMPLE(0.01)',
    search_params={"metric_type": "L2", "params": {}},
    output_fields=["title", "author", "price"],
    limit=10
)

print(f"Found {len(search_results[0])} similar books in sample")

Java

import io.milvus.v2.service.vector.request.SearchReq
import io.milvus.v2.service.vector.request.data.FloatVec;
import io.milvus.v2.service.vector.response.SearchResp

FloatVec queryVector = new FloatVec(new float[]{0.1f, 0.2f, 0.3f, 0.4f, 0.5f});
SearchReq searchReq = SearchReq.builder()
        .collectionName("product_catalog")
        .data(Collections.singletonList(queryVector))
        .topK(10)
        .filter("category == \"books\" AND RANDOM_SAMPLE(0.01)")
        .outputFields(Arrays.asList("title", "author", "price"))
        .build();

SearchResp searchResp = client.search(searchReq);

List<List<SearchResp.SearchResult>> searchResults = searchResp.getSearchResults();
for (List<SearchResp.SearchResult> results : searchResults) {
    System.out.println("TopK results:");
    for (SearchResp.SearchResult result : results) {
        System.out.println(result);
    }
}

Go

queryVector := []float32{0.1, 0.2, 0.3, 0.4, 0.5}

resultSets, err := client.Search(ctx, milvusclient.NewSearchOption(
    "product_catalog", // collectionName
    10,               // limit
    []entity.Vector{entity.FloatVector(queryVector)},
).WithConsistencyLevel(entity.ClStrong).
    WithFilter("category == \"books\" AND RANDOM_SAMPLE(0.01)").
    WithOutputFields("title", "author", "price"))
if err != nil {
    fmt.Println(err.Error())
    // handle error
}

for _, resultSet := range resultSets {
    fmt.Println("title: ", resultSet.GetColumn("title").FieldData().GetScalars())
    fmt.Println("author: ", resultSet.GetColumn("author").FieldData().GetScalars())
    fmt.Println("price: ", resultSet.GetColumn("price").FieldData().GetScalars())
}

NodeJS

// node

cURL

# restful

最佳实践

• 从小处入手： 从较小的采样因子（0.001-0.01）开始进行初步探索

• 开发工作流程：在开发过程中使用采样，在生产查询中移除采样

• 统计有效性：较大的样本可提供更准确的统计表示

• 性能测试：监控查询性能，并根据需要调整取样系数