最近邻查询¶

最近邻查询语句用于在图中搜索向量的最近邻。该语句支持两种搜索类型：K 最近邻（k-nearest neighbor，简称 KNN）搜索和近似最近邻（Approximate Nearest Neighbor，简称 ANN）搜索。

KNN 是一种精确搜索，返回查询向量的 K 个最近邻。它不依赖于向量索引，适用于小型的图和低维度的向量。

ANN 是一种近似搜索，返回查询向量的近似最近邻。它依赖于向量索引，适用于大型的图和高维度的向量。

在执行 ANN 搜索前，需要在待搜索的向量上预先创建向量索引；否则，ANN 搜索将自动转换为 KNN 搜索。更多信息，请参见 CREATE VECTOR INDEX。

语法¶

语法图源码

knn_search_statement
ann_search_statement
ann_search_option

knn_search_statement ::= order_by_clause limit_clause 

ann_search_statement ::= order_by_clause ('APPROX'| 'APPROXIMATE') limit_clause ('OPTIONS' '{' ann_search_option (',' ann_search_option)* '}' )

ann_search_option::=  
    'METRIC' ':' ( 'L2' | 'IP' )
  | 'TYPE' ':' ( 'IVF' | 'HNSW' )
  | 'NPROBE' ':' int_value
  | 'EFSEARCH' ':' int_value

参数¶

`order_by_clause`¶

基于查询向量和索引向量之间的相似度对搜索结果进行排序的子句。该子句必须包含一个用于测量向量相似度的函数。

支持以下函数：

euclidean()：若在 ANN 搜索中使用，需满足以下条件：
- 已为待搜索的向量创建度量为L2的向量索引；否则，ANN 搜索会自动转换为 KNN 搜索。
- 搜索结果按升序排序。
inner_product()：若在 ANN 搜索中使用，需满足以下条件：
- 已为待搜索的向量创建度量为IP的向量索引；否则，ANN 搜索会自动转换为 KNN 搜索。
- 搜索结果按降序排序。
cosine()：若使用该函数，则仅支持 KNN 搜索。

更多信息，请参见 ORDER BY。

`limit_clause`¶

限定返回的搜索结果数量的子句。更多信息，请参见 LIMIT。

`ann_search_option`¶

执行 ANN 搜索的选项。

`METRIC`¶

用于衡量查询向量和索引向量之间相似度的度量。该值会根据前面order_by_clause中指定的函数自动设置，规则如下：

若函数为euclidean()，则度量默认设置为L2，即根据向量之间的欧几里得距离计算相似度。更多信息，请参见 euclidean()。
若函数为inner_product()，则度量默认设置为IP，即根据向量之间的内积计算相似度。更多信息，请参见 inner_product()。

若此处指定的值与前面order_by_clause中指定的函数相矛盾，则会出现报错提醒。

`TYPE`¶

用于执行 ANN 搜索的向量索引类型。有效值为IVF和HNSW。若未指定此选项，则默认值IVF生效。更多信息，请参见 CREATE VECTOR INDEX。

`NPROBE`¶

在 IVF 搜索中搜索的聚类数量。该选项仅在选项TYPE设置为IVF时生效。

该选项用于权衡 IVF 算法的搜索速度和搜索准确度。取值越大，搜索准确度越高，但搜索速度越慢。

有效值范围为1到NLIST，其中NLIST为创建 IVF 索引时创建的聚类数量。更多信息，请参见 NLIST。

若NPROBE设置为与NLIST相同的取值，则理论上搜索的召回率最高，但搜索速度最慢。

若未指定该选项且TYPE设置为IVF，则默认值8生效。

`EFSEARCH`¶

在 HNSW 搜索中动态候选列表的大小，该列表大小决定了搜索过程中评估的邻居数量。该选项仅在选项TYPE设置为HNSW时生效。

该选项用于权衡 HNSW 算法的搜索速度和搜索准确度。取值越大，搜索准确度越高，但搜索速度越慢。

有效值范围为1到无穷大。

若未指定该选项且TYPE设置为HNSW，则默认值16生效。

示例¶

以下示例逐步介绍了如何在图中执行 KNN 和 ANN 搜索：

基于 CREATE GRAPH TYPE 中创建的图类型ann_test创建图。

CREATE GRAPH IF NOT EXISTS ann_test_graph TYPED ann_test

往图ann_test_graph中插入点和边。

'''
// 往图中插入 27 个点类型为`N1`的点。
TABLE t {id, vec1, vec2, vec3} =   
    {id:1, vec1:VECTOR<3,float>([1.0, 2.0, 3.0]), vec2:VECTOR<7,float>([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0]), vec3:VECTOR<3,float>([1.0, 2.0, 3.0])}, 
    {id:2, vec1:VECTOR<3,float>([2.0, 3.0, 4.0]), vec2:VECTOR<7,float>([2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0]), vec3:VECTOR<3,float>([2.0, 3.0, 4.0])}, 
    {id:3, vec1:VECTOR<3,float>([3.0, 4.0, 5.0]), vec2:VECTOR<7,float>([3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0]), vec3:VECTOR<3,float>([3.0, 4.0, 5.0])}, 
    {id:4, vec1:VECTOR<3,float>([4.0, 5.0, 6.0]), vec2:VECTOR<7,float>([4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0]), vec3:VECTOR<3,float>([4.0, 5.0, 6.0])}, 
    {id:5, vec1:VECTOR<3,float>([5.0, 6.0, 7.0]), vec2:VECTOR<7,float>([5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0]), vec3:VECTOR<3,float>([5.0, 6.0, 7.0])}, 
    {id:6, vec1:VECTOR<3,float>([6.0, 7.0, 8.0]), vec2:VECTOR<7,float>([6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0]), vec3:VECTOR<3,float>([6.0, 7.0, 8.0])}, 
    {id:7, vec1:VECTOR<3,float>([7.0, 8.0, 9.0]), vec2:VECTOR<7,float>([7.0, 8.0, 9.0, 10.0, 11.0, 12.0, 13.0]), vec3:VECTOR<3,float>([7.0, 8.0, 9.0])}, 
    {id:8, vec1:VECTOR<3,float>([8.0, 9.0, 10.0]), vec2:VECTOR<7,float>([8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0]), vec3:VECTOR<3,float>([8.0, 9.0, 10.0])}, 
    {id:9, vec1:VECTOR<3,float>([9.0, 10.0, 11.0]), vec2:VECTOR<7,float>([9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0]), vec3:VECTOR<3,float>([9.0, 10.0, 11.0])}, 
    {id:10, vec1:VECTOR<3,float>([10.0, 11.0, 12.0]), vec2:VECTOR<7,float>([10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0]), vec3:VECTOR<3,float>([10.0, 11.0, 12.0])}, 
    {id:11, vec1:VECTOR<3,float>([11.0, 12.0, 13.0]), vec2:VECTOR<7,float>([11.0, 12.0, 13.0, 14.0, 15.0, 16.0, 17.0]), vec3:VECTOR<3,float>([11.0, 12.0, 13.0])}, 
    {id:12, vec1:VECTOR<3,float>([12.0, 13.0, 14.0]), vec2:VECTOR<7,float>([12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0]), vec3:VECTOR<3,float>([12.0, 13.0, 14.0])}, 
    {id:13, vec1:VECTOR<3,float>([13.0, 14.0, 15.0]), vec2:VECTOR<7,float>([13.0, 14.0, 15.0, 16.0, 17.0, 18.0, 19.0]), vec3:VECTOR<3,float>([13.0, 14.0, 15.0])}, 
    {id:14, vec1:VECTOR<3,float>([14.0, 15.0, 16.0]), vec2:VECTOR<7,float>([14.0, 15.0, 16.0, 17.0, 18.0, 19.0, 20.0]), vec3:VECTOR<3,float>([14.0, 15.0, 16.0])}, 
    {id:15, vec1:VECTOR<3,float>([15.0, 16.0, 17.0]), vec2:VECTOR<7,float>([15.0, 16.0, 17.0, 18.0, 19.0, 20.0, 21.0]), vec3:VECTOR<3,float>([15.0, 16.0, 17.0])}, 
    {id:16, vec1:VECTOR<3,float>([16.0, 17.0, 18.0]), vec2:VECTOR<7,float>([16.0, 17.0, 18.0, 19.0, 20.0, 21.0, 22.0]), vec3:VECTOR<3,float>([16.0, 17.0, 18.0])}, 
    {id:17, vec1:VECTOR<3,float>([17.0, 18.0, 19.0]), vec2:VECTOR<7,float>([17.0, 18.0, 19.0, 20.0, 21.0, 22.0, 23.0]), vec3:VECTOR<3,float>([17.0, 18.0, 19.0])}, 
    {id:18, vec1:VECTOR<3,float>([18.0, 19.0, 20.0]), vec2:VECTOR<7,float>([18.0, 19.0, 20.0, 21.0, 22.0, 23.0, 24.0]), vec3:VECTOR<3,float>([18.0, 19.0, 20.0])}, 
    {id:19, vec1:VECTOR<3,float>([19.0, 20.0, 21.0]), vec2:VECTOR<7,float>([19.0, 20.0, 21.0, 22.0, 23.0, 24.0, 25.0]), vec3:VECTOR<3,float>([19.0, 20.0, 21.0])}, 
    {id:20, vec1:VECTOR<3,float>([20.0, 21.0, 22.0]), vec2:VECTOR<7,float>([20.0, 21.0, 22.0, 23.0, 24.0, 25.0, 26.0]), vec3:VECTOR<3,float>([20.0, 21.0, 22.0])}, 
    {id:21, vec1:VECTOR<3,float>([21.0, 22.0, 23.0]), vec2:VECTOR<7,float>([21.0, 22.0, 23.0, 24.0, 25.0, 26.0, 27.0]), vec3:VECTOR<3,float>([21.0, 22.0, 23.0])}, 
    {id:22, vec1:VECTOR<3,float>([22.0, 23.0, 24.0]), vec2:VECTOR<7,float>([22.0, 23.0, 24.0, 25.0, 26.0, 27.0, 28.0]), vec3:VECTOR<3,float>([22.0, 23.0, 24.0])}, 
    {id:23, vec1:VECTOR<3,float>([23.0, 24.0, 25.0]), vec2:VECTOR<7,float>([23.0, 24.0, 25.0, 26.0, 27.0, 28.0, 29.0]), vec3:VECTOR<3,float>([23.0, 24.0, 25.0])}, 
    {id:24, vec1:VECTOR<3,float>([24.0, 25.0, 26.0]), vec2:VECTOR<7,float>([24.0, 25.0, 26.0, 27.0, 28.0, 29.0, 30.0]), vec3:VECTOR<3,float>([24.0, 25.0, 26.0])}, 
    {id:25, vec1:VECTOR<3,float>([25.0, 26.0, 27.0]), vec2:VECTOR<7,float>([25.0, 26.0, 27.0, 28.0, 29.0, 30.0, 31.0]), vec3:VECTOR<3,float>([25.0, 26.0, 27.0])}, 
    {id:26, vec1:VECTOR<3,float>([26.0, 27.0, 28.0]), vec2:VECTOR<7,float>([26.0, 27.0, 28.0, 29.0, 30.0, 31.0, 32.0]), vec3:VECTOR<3,float>([26.0, 27.0, 28.0])}, 
    {id:27, vec1:VECTOR<3,float>([27.0, 28.0, 29.0]), vec2:VECTOR<7,float>([27.0, 28.0, 29.0, 30.0, 31.0, 32.0, 33.0]), vec3:VECTOR<3,float>([27.0, 28.0, 29.0])} 
USE ann_test_graph 
FOR r IN t 
INSERT OR IGNORE (@N1{id:r.id,vec1:r.vec1,vec2:r.vec2,vec3:r.vec3})
'''

'''
// 往图中插入 19 个点类型为`N2`的点。
TABLE t {id, vec1, vec2, vec5} = 
    {id:1, vec1:VECTOR<3,float>([1.1, 2.1, 3.1]), vec2:VECTOR<7,float>([1.2, 2.2, 3.2, 4.2, 5.2, 6.2, 7.2]), vec5:VECTOR<5,float>([1.3, 2.3, 3.3, 4.3, 5.3])}, 
    {id:2, vec1:VECTOR<3,float>([2.1, 3.1, 4.1]), vec2:VECTOR<7,float>([2.2, 3.2, 4.2, 5.2, 6.2, 7.2, 8.2]), vec5:VECTOR<5,float>([2.3, 3.3, 4.3, 5.3, 6.3])}, 
    {id:3, vec1:VECTOR<3,float>([3.1, 4.1, 5.1]), vec2:VECTOR<7,float>([3.2, 4.2, 5.2, 6.2, 7.2, 8.2, 9.2]), vec5:VECTOR<5,float>([3.3, 4.3, 5.3, 6.3, 7.3])}, 
    {id:4, vec1:VECTOR<3,float>([4.1, 5.1, 6.1]), vec2:VECTOR<7,float>([4.2, 5.2, 6.2, 7.2, 8.2, 9.2, 10.2]), vec5:VECTOR<5,float>([4.3, 5.3, 6.3, 7.3, 8.3])}, 
    {id:5, vec1:VECTOR<3,float>([5.1, 6.1, 7.1]), vec2:VECTOR<7,float>([5.2, 6.2, 7.2, 8.2, 9.2, 10.2, 11.2]), vec5:VECTOR<5,float>([5.3, 6.3, 7.3, 8.3, 9.3])}, 
    {id:6, vec1:VECTOR<3,float>([6.1, 7.1, 8.1]), vec2:VECTOR<7,float>([6.2, 7.2, 8.2, 9.2, 10.2, 11.2, 12.2]), vec5:VECTOR<5,float>([6.3, 7.3, 8.3, 9.3, 10.3])}, 
    {id:7, vec1:VECTOR<3,float>([7.1, 8.1, 9.1]), vec2:VECTOR<7,float>([7.2, 8.2, 9.2, 10.2, 11.2, 12.2, 13.2]), vec5:VECTOR<5,float>([7.3, 8.3, 9.3, 10.3, 11.3])}, 
    {id:8, vec1:VECTOR<3,float>([8.1, 9.1, 10.1]), vec2:VECTOR<7,float>([8.2, 9.2, 10.2, 11.2, 12.2, 13.2, 14.2]), vec5:VECTOR<5,float>([8.3, 9.3, 10.3, 11.3, 12.3])}, 
    {id:9, vec1:VECTOR<3,float>([9.1, 10.1, 11.1]), vec2:VECTOR<7,float>([9.2, 10.2, 11.2, 12.2, 13.2, 14.2, 15.2]), vec5:VECTOR<5,float>([9.3, 10.3, 11.3, 12.3, 13.3])}, 
    {id:10, vec1:VECTOR<3,float>([10.1, 11.1, 12.1]), vec2:VECTOR<7,float>([10.2, 11.2, 12.2, 13.2, 14.2, 15.2, 16.2]), vec5:VECTOR<5,float>([10.3, 11.3, 12.3, 13.3, 14.3])}, 
    {id:11, vec1:VECTOR<3,float>([11.1, 12.1, 13.1]), vec2:VECTOR<7,float>([11.2, 12.2, 13.2, 14.2, 15.2, 16.2, 17.2]), vec5:VECTOR<5,float>([11.3, 12.3, 13.3, 14.3, 15.3])}, 
    {id:12, vec1:VECTOR<3,float>([12.1, 13.1, 14.1]), vec2:VECTOR<7,float>([12.2, 13.2, 14.2, 15.2, 16.2, 17.2, 18.2]), vec5:VECTOR<5,float>([12.3, 13.3, 14.3, 15.3, 16.3])}, 
    {id:13, vec1:VECTOR<3,float>([13.1, 14.1, 15.1]), vec2:VECTOR<7,float>([13.2, 14.2, 15.2, 16.2, 17.2, 18.2, 19.2]), vec5:VECTOR<5,float>([13.3, 14.3, 15.3, 16.3, 17.3])}, 
    {id:14, vec1:VECTOR<3,float>([14.1, 15.1, 16.1]), vec2:VECTOR<7,float>([14.2, 15.2, 16.2, 17.2, 18.2, 19.2, 20.2]), vec5:VECTOR<5,float>([14.3, 15.3, 16.3, 17.3, 18.3])}, 
    {id:15, vec1:VECTOR<3,float>([15.1, 16.1, 17.1]), vec2:VECTOR<7,float>([15.2, 16.2, 17.2, 18.2, 19.2, 20.2, 21.2]), vec5:VECTOR<5,float>([15.3, 16.3, 17.3, 18.3, 19.3])}, 
    {id:16, vec1:VECTOR<3,float>([16.1, 17.1, 18.1]), vec2:VECTOR<7,float>([16.2, 17.2, 18.2, 19.2, 20.2, 21.2, 22.2]), vec5:VECTOR<5,float>([16.3, 17.3, 18.3, 19.3, 20.3])}, 
    {id:17, vec1:VECTOR<3,float>([17.1, 18.1, 19.1]), vec2:VECTOR<7,float>([17.2, 18.2, 19.2, 20.2, 21.2, 22.2, 23.2]), vec5:VECTOR<5,float>([17.3, 18.3, 19.3, 20.3, 21.3])}, 
    {id:18, vec1:VECTOR<3,float>([18.1, 19.1, 20.1]), vec2:VECTOR<7,float>([18.2, 19.2, 20.2, 21.2, 22.2, 23.2, 24.2]), vec5:VECTOR<5,float>([18.3, 19.3, 20.3, 21.3, 22.3])}, 
    {id:19, vec1:VECTOR<3,float>([19.1, 20.1, 21.1]), vec2:VECTOR<7,float>([19.2, 20.2, 21.2, 22.2, 23.2, 24.2, 25.2]), vec5:VECTOR<5,float>([19.3, 20.3, 21.3, 22.3, 23.3])} 
USE ann_test_graph 
FOR r IN t 
INSERT OR IGNORE (@N2{id:r.id,vec1:r.vec1,vec2:r.vec2,vec5:r.vec5})
'''

'''
// 往图中插入 27 条边类型为`E1`的边。
TABLE t {src, dst, id, vec1, vec2, vec3} = 
    {src: 1, dst:1, id:1, vec1:VECTOR<3,float>([1.11, 2.12, 3.13]), vec2:VECTOR<7,float>([1.21, 2.22, 3.23, 4.24, 5.25, 6.26, 7.27]), vec3:VECTOR<3,float>([1.31, 2.32, 3.33])}, 
    {src: 2, dst:2, id:2, vec1:VECTOR<3,float>([2.11, 3.12, 4.13]), vec2:VECTOR<7,float>([2.21, 3.22, 4.23, 5.24, 6.25, 7.26, 8.27]), vec3:VECTOR<3,float>([2.31, 3.32, 4.33])}, 
    {src: 3, dst:3, id:3, vec1:VECTOR<3,float>([3.11, 4.12, 5.13]), vec2:VECTOR<7,float>([3.21, 4.22, 5.23, 6.24, 7.25, 8.26, 9.27]), vec3:VECTOR<3,float>([3.31, 4.32, 5.33])}, 
    {src: 4, dst:4, id:4, vec1:VECTOR<3,float>([4.11, 5.12, 6.13]), vec2:VECTOR<7,float>([4.21, 5.22, 6.23, 7.24, 8.25, 9.26, 10.27]), vec3:VECTOR<3,float>([4.31, 5.32, 6.33])}, 
    {src: 5, dst:5, id:5, vec1:VECTOR<3,float>([5.11, 6.12, 7.13]), vec2:VECTOR<7,float>([5.21, 6.22, 7.23, 8.24, 9.25, 10.26, 11.27]), vec3:VECTOR<3,float>([5.31, 6.32, 7.33])}, 
    {src: 6, dst:6, id:6, vec1:VECTOR<3,float>([6.11, 7.12, 8.13]), vec2:VECTOR<7,float>([6.21, 7.22, 8.23, 9.24, 10.25, 11.26, 12.27]), vec3:VECTOR<3,float>([6.31, 7.32, 8.33])}, 
    {src: 7, dst:7, id:7, vec1:VECTOR<3,float>([7.11, 8.12, 9.13]), vec2:VECTOR<7,float>([7.21, 8.22, 9.23, 10.24, 11.25, 12.26, 13.27]), vec3:VECTOR<3,float>([7.31, 8.32, 9.33])}, 
    {src: 8, dst:8, id:8, vec1:VECTOR<3,float>([8.11, 9.12, 10.13]), vec2:VECTOR<7,float>([8.21, 9.22, 10.23, 11.24, 12.25, 13.26, 14.27]), vec3:VECTOR<3,float>([8.31, 9.32, 10.33])}, 
    {src: 9, dst:9, id:9, vec1:VECTOR<3,float>([9.11, 10.12, 11.13]), vec2:VECTOR<7,float>([9.21, 10.22, 11.23, 12.24, 13.25, 14.26, 15.27]), vec3:VECTOR<3,float>([9.31, 10.32, 11.33])}, 
    {src: 10, dst:10, id:10, vec1:VECTOR<3,float>([10.11, 11.12, 12.13]), vec2:VECTOR<7,float>([10.21, 11.22, 12.23, 13.24, 14.25, 15.26, 16.27]), vec3:VECTOR<3,float>([10.31, 11.32, 12.33])}, 
    {src: 11, dst:11, id:11, vec1:VECTOR<3,float>([11.11, 12.12, 13.13]), vec2:VECTOR<7,float>([11.21, 12.22, 13.23, 14.24, 15.25, 16.26, 17.27]), vec3:VECTOR<3,float>([11.31, 12.32, 13.33])}, 
    {src: 12, dst:12, id:12, vec1:VECTOR<3,float>([12.11, 13.12, 14.13]), vec2:VECTOR<7,float>([12.21, 13.22, 14.23, 15.24, 16.25, 17.26, 18.27]), vec3:VECTOR<3,float>([12.31, 13.32, 14.33])}, 
    {src: 13, dst:13, id:13, vec1:VECTOR<3,float>([13.11, 14.12, 15.13]), vec2:VECTOR<7,float>([13.21, 14.22, 15.23, 16.24, 17.25, 18.26, 19.27]), vec3:VECTOR<3,float>([13.31, 14.32, 15.33])}, 
    {src: 14, dst:14, id:14, vec1:VECTOR<3,float>([14.11, 15.12, 16.13]), vec2:VECTOR<7,float>([14.21, 15.22, 16.23, 17.24, 18.25, 19.26, 20.27]), vec3:VECTOR<3,float>([14.31, 15.32, 16.33])}, 
    {src: 15, dst:15, id:15, vec1:VECTOR<3,float>([15.11, 16.12, 17.13]), vec2:VECTOR<7,float>([15.21, 16.22, 17.23, 18.24, 19.25, 20.26, 21.27]), vec3:VECTOR<3,float>([15.31, 16.32, 17.33])}, 
    {src: 16, dst:16, id:16, vec1:VECTOR<3,float>([16.11, 17.12, 18.13]), vec2:VECTOR<7,float>([16.21, 17.22, 18.23, 19.24, 20.25, 21.26, 22.27]), vec3:VECTOR<3,float>([16.31, 17.32, 18.33])}, 
    {src: 17, dst:17, id:17, vec1:VECTOR<3,float>([17.11, 18.12, 19.13]), vec2:VECTOR<7,float>([17.21, 18.22, 19.23, 20.24, 21.25, 22.26, 23.27]), vec3:VECTOR<3,float>([17.31, 18.32, 19.33])}, 
    {src: 18, dst:18, id:18, vec1:VECTOR<3,float>([18.11, 19.12, 20.13]), vec2:VECTOR<7,float>([18.21, 19.22, 20.23, 21.24, 22.25, 23.26, 24.27]), vec3:VECTOR<3,float>([18.31, 19.32, 20.33])}, 
    {src: 19, dst:19, id:19, vec1:VECTOR<3,float>([19.11, 20.12, 21.13]), vec2:VECTOR<7,float>([19.21, 20.22, 21.23, 22.24, 23.25, 24.26, 25.27]), vec3:VECTOR<3,float>([19.31, 20.32, 21.33])}, 
    {src: 20, dst:17, id:20, vec1:VECTOR<3,float>([20.11, 21.12, 22.13]), vec2:VECTOR<7,float>([20.21, 21.22, 22.23, 23.24, 24.25, 25.26, 26.27]), vec3:VECTOR<3,float>([20.31, 21.32, 22.33])}, 
    {src: 21, dst:17, id:21, vec1:VECTOR<3,float>([21.11, 22.12, 23.13]), vec2:VECTOR<7,float>([21.21, 22.22, 23.23, 24.24, 25.25, 26.26, 27.27]), vec3:VECTOR<3,float>([21.31, 22.32, 23.33])}, 
    {src: 22, dst:17, id:22, vec1:VECTOR<3,float>([22.11, 23.12, 24.13]), vec2:VECTOR<7,float>([22.21, 23.22, 24.23, 25.24, 26.25, 27.26, 28.27]), vec3:VECTOR<3,float>([22.31, 23.32, 24.33])}, 
    {src: 23, dst:17, id:23, vec1:VECTOR<3,float>([23.11, 24.12, 25.13]), vec2:VECTOR<7,float>([23.21, 24.22, 25.23, 26.24, 27.25, 28.26, 29.27]), vec3:VECTOR<3,float>([23.31, 24.32, 25.33])}, 
    {src: 24, dst:17, id:24, vec1:VECTOR<3,float>([24.11, 25.12, 26.13]), vec2:VECTOR<7,float>([24.21, 25.22, 26.23, 27.24, 28.25, 29.26, 30.27]), vec3:VECTOR<3,float>([24.31, 25.32, 26.33])}, 
    {src: 25, dst:17, id:25, vec1:VECTOR<3,float>([25.11, 26.12, 27.13]), vec2:VECTOR<7,float>([25.21, 26.22, 27.23, 28.24, 29.25, 30.26, 31.27]), vec3:VECTOR<3,float>([25.31, 26.32, 27.33])}, 
    {src: 26, dst:17, id:26, vec1:VECTOR<3,float>([26.11, 27.12, 28.13]), vec2:VECTOR<7,float>([26.21, 27.22, 28.23, 29.24, 30.25, 31.26, 32.27]), vec3:VECTOR<3,float>([26.31, 27.32, 28.33])}, 
    {src: 27, dst:17, id:27, vec1:VECTOR<3,float>([27.11, 28.12, 29.13]), vec2:VECTOR<7,float>([27.21, 28.22, 29.23, 30.24, 31.25, 32.26, 33.27]), vec3:VECTOR<3,float>([27.31, 28.32, 29.33])} 
USE ann_test_graph 
FOR r IN t 
MATCH (n1:N1{id:r.src}), (n2:N2{id:r.dst}) 
INSERT OR IGNORE (n1)-[@E1{id:r.id,vec1:r.vec1,vec2:r.vec2,vec3:r.vec3}]->(n2)
'''

在图ann_test_graph上创建向量索引。

'''
// 基于度量`L2`，为点类型`N1`的向量属性`vec1`创建 IVF 向量索引。
USE ann_test_graph CREATE VECTOR INDEX ann_node_ivf_l2 ON NODE N1::vec1 OPTIONS {DIM: 3, METRIC: L2, TYPE: IVF, NLIST: 10, TRAINSIZE: 4 }
'''

'''
// 基于度量`IP`，为点类型`N1`的向量属性`vec1`创建 HNSW 向量索引。
USE ann_test_graph CREATE VECTOR INDEX ann_node_hnsw_ip ON NODE N1::vec1 OPTIONS {DIM: 3, METRIC: IP, TYPE: HNSW, MAXDEGREE: 3, EFCONSTRUCTION: 6, CAPACITY: 500}
'''

'''
// 基于度量`L2`，为边类型`E1`的向量属性`vec1`创建 IVF 向量索引。
USE ann_test_graph CREATE VECTOR INDEX ann_edge_ivf_l2 ON EDGE E1::vec1 OPTIONS {DIM: 3, METRIC: L2, TYPE: IVF, NLIST: 10, TRAINSIZE: 4}
'''

'''
// 基于度量`IP`，为边类型`E1`的向量属性`vec1`创建 HNSW 向量索引。
USE ann_test_graph CREATE VECTOR INDEX ann_edge_hnsw_ip ON EDGE E1::vec1 OPTIONS {DIM: 3, METRIC: IP, TYPE: HNSW, MAXDEGREE: 3, EFCONSTRUCTION: 6, CAPACITY: 500}
'''

查询图ann_test_graph以查找给定向量的最近邻。

执行 KNN 搜索并根据欧几里得距离查找最近邻

'''
USE ann_test_graph 
MATCH (v:N1) 
ORDER BY euclidean(v.vec1, VECTOR<3,float>([1.0, 2.0, 3.0])) ASC 
LIMIT 3 
RETURN v.id AS vid, labels(v) AS vlabel, v.vec1 AS vec1, euclidean(v.vec1, VECTOR<3,float>([1.0, 2.0, 3.0])) AS distance
'''

返回：

+-----+--------+---------------+-----------+
| vid | vlabel | vec1          | distance  |
+-----+--------+---------------+-----------+
| 1   | [N1]   | [1.0,2.0,3.0] | 0.0       |
| 2   | [N1]   | [2.0,3.0,4.0] | 1.7320508 |
| 3   | [N1]   | [3.0,4.0,5.0] | 3.4641016 |
+-----+--------+---------------+-----------+

执行 KNN 搜索并根据内积查找最近邻

'''
USE ann_test_graph 
MATCH (v:N1) 
ORDER BY inner_product(v.vec1, VECTOR<3,float>([1.0, 2.0, 3.0])) DESC 
LIMIT 3 
RETURN v.id AS vid, labels(v) AS vlabel, v.vec1 AS vec1, inner_product(v.vec1, VECTOR<3,float>([1.0, 2.0, 3.0])) AS distance
'''

返回：

+-----+--------+------------------+----------+
| vid | vlabel | vec1             | distance |
+-----+--------+------------------+----------+
| 27  | [N1]   | [27.0,28.0,29.0] | 170.0    |
| 26  | [N1]   | [26.0,27.0,28.0] | 164.0    |
| 25  | [N1]   | [25.0,26.0,27.0] | 158.0    |
+-----+--------+------------------+----------+

执行 ANN 搜索并根据欧几里得距离查找最近邻

'''
USE ann_test_graph 
MATCH (v:N1) 
ORDER BY euclidean(v.vec1, VECTOR<3,float>([1.0, 2.0, 3.0])) ASC 
APPROXIMATE 
LIMIT 3 
OPTIONS { METRIC: L2, TYPE: IVF, NPROBE: 8 } 
RETURN v.id AS vid, labels(v) AS vlabel, v.vec1 AS vec1, euclidean(v.vec1, VECTOR<3,float>([1.0, 2.0, 3.0])) AS distance
'''

返回：

+-----+--------+---------------+-----------+
| vid | vlabel | vec1          | distance  |
+-----+--------+---------------+-----------+
| 1   | [N1]   | [1.0,2.0,3.0] | 0.0       |
| 2   | [N1]   | [2.0,3.0,4.0] | 1.7320508 |
| 3   | [N1]   | [3.0,4.0,5.0] | 3.4641016 |
+-----+--------+---------------+-----------+

执行 ANN 搜索并根据内积查找最近邻

'''
USE ann_test_graph 
MATCH (v:N1) 
ORDER BY inner_product(v.vec1, VECTOR<3,float>([1.0, 2.0, 3.0])) DESC 
APPROXIMATE 
LIMIT 3 
OPTIONS { METRIC: IP, TYPE: HNSW, EFSEARCH: 8 } 
RETURN v.id AS vid, labels(v) AS vlabel, v.vec1 AS vec1, inner_product(v.vec1, VECTOR<3,  float>([1.0, 2.0, 3.0])) AS distance
'''

返回：

+-----+--------+------------------+----------+
| vid | vlabel | vec1             | distance |
+-----+--------+------------------+----------+
| 27  | [N1]   | [27.0,28.0,29.0] | 170.0    |
| 26  | [N1]   | [26.0,27.0,28.0] | 164.0    |
| 25  | [N1]   | [25.0,26.0,27.0] | 158.0    |
+-----+--------+------------------+----------+

最近邻查询¶

语法¶

参数¶

order_by_clause¶

limit_clause¶

ann_search_option¶

METRIC¶

TYPE¶

NPROBE¶

EFSEARCH¶

示例¶