Rate This Document
Findability
Accuracy
Completeness
Readability

Parametrized Constructor

API Definition

KBest(int K_features, int R, int L, int A, const char *metric, const std::string &init_builder_type);

KBest(int K_features, int R, int L, float A, int graph_opt_iter, const std::string &metric, const std::string &init_builder_type, const std::string &index_type);

Function

Initializes the build of graph search.

Parameters

Parameter

Description

Data Type

Value Range

K_features

Feature dimension.

int

[1, 2999]

R

Number of neighboring nodes.

int

[11, 499]

L

Size of the candidate node list during the graph build.

int

[11, 1999]

A

Angle threshold during the pruning of graph build.

int

[1, 360]

metric

Distance measurement indicator.

const char *

L2 or IP.

  • L2 indicates the Euclidean distance.
  • IP indicates the inner product distance.

init_builder_type

Algorithm for building the k-nearest neighbor graph.

const std::string &

RNNDescent and NNDescent indicate two different algorithms for building the k-nearest neighbor graph. RNNDescent is often used.

graph_opt_iter

Number of rounds for graph index self-iteration.

int

[0, 30]

index_type

Neighbor selection strategy.

const std::string &

HNSW, NSG, TSDG, and SSG indicate the four different neighbor selection strategies.