Package 'RNGT'

Title: Wrappers for 'NGT'
Description: Wrappers for 'NGT' (Neighborhood Graph and Tree for indexing high-dimensional data) which performs high-speed approximate nearest neighbor searches against a large volume of data in high dimensional vector data space.
Authors: Akiru Kato [aut, cre]
Maintainer: Akiru Kato <[email protected]>
License: Apache License (>= 2)
Version: 0.0.0.9001
Built: 2024-09-02 04:53:56 UTC
Source: https://github.com/paithiov909/RNGT

Help Index

GloVe Model of Pokémon

Description

A GloVe model that trained on 11,446 Pokémon parties (for single battles in 2018), while considering each party as a sentence and each Pokémon as a word, and pruning vocabularies.

Usage

gen7singles2018

Format

An object of class tbl_df (inherits from tbl, data.frame) with 181 rows and 66 columns.

R6 Class for Graph and Tree Based Index

Description

Graph and tree based index.

Public fields

path

path of the index.

Methods

Public methods

Method new()

Creates a new NgtIndex object.

Usage
NgtIndex$new(path, sub_dir = rand_name("NgtIndex"))
Arguments
path

path to the NGT index.

sub_dir

sub-directory of the index.

Returns

a new NgtIndex object.

Method open()

Opens a NGT index.

Usage
NgtIndex$open(read_only = FALSE, tree_disabled = FALSE, log_disabled = FALSE)
Arguments
read_only

whether the index is read only.

tree_disabled

whether the tree is disabled.

log_disabled

whether the log is disabled.

Method create()

Creates an empty index with the specified parameters.

Usage
NgtIndex$create(
  dimension,
  edge_size_for_creation = 10,
  edge_size_for_search = 40,
  distance_type = c("l2", "l1", "normalized_l2", "hamming", "jaccard", "sparse_jaccard",
    "angle", "normalized_angle", "cosine", "normalized_cosine", "normalized_l2"),
  object_type = c("float", "byte", "float16")
)
Arguments
dimension

dimension of the vectors.

edge_size_for_creation

number of edges for each node in the graph.

edge_size_for_search

number of edges to search.

distance_type

distance type.

object_type

object type.

sub_dir

sub directory to store the index in.

Method search()

Searches for the k approximate nearest neighbors of the specifiecd query object.

Usage
NgtIndex$search(
  query,
  k = 20L,
  epsilon = 0.1,
  edge_size = -1L,
  expected_accuracy = -1,
  with_distance = TRUE
)
Arguments
query

query object.

k

number of nearest neighbors.

epsilon

epsilon which defines the explored range for the graph

edge_size

number of edges for each node to explore the graph

expected_accuracy

expected accuracy.

with_distance

whether to return distance.

Returns

tibble.

Method save()

Saves the index.

Usage
NgtIndex$save(path)
Arguments
path

path to save the index. defaults to the path of the index.

Method remove()

Removes objects from the index by their IDs.

Usage
NgtIndex$remove(ids)
Arguments
ids

IDs of the objects to be removed.

Returns

integers; ids is returned invisibly as is.

Method refine_anng()

Refines the index with the specified parameters.

Usage
NgtIndex$refine_anng(
  epsilon,
  accuracy,
  num_edges,
  num_edges_for_search,
  batch_size
)
Arguments
epsilon

epsilon which defines the explored range for the graph

accuracy

expected accuracy.

num_edges

number of edges for each node to explore the graph

num_edges_for_search

number of edges to search.

batch_size

batch size.

Method get_object()

Gets objects from the index by their IDs.

Usage
NgtIndex$get_object(ids)
Arguments
ids

IDs of the objects to be retrieved.

Returns

tibble.

Method reset_defaults()

Resets the default parameters of the index.

Usage
NgtIndex$reset_defaults(
  num_of_search_objects,
  search_radius,
  epsilon,
  edge_size,
  expected_accuracy
)
Arguments
num_of_search_objects

number of search objects.

search_radius

search radius.

epsilon

epsilon.

edge_size

edge size.

expected_accuracy

expected accuracy.

Method get_info()

Gets information of the index.

Usage
NgtIndex$get_info()
Returns

named numeric vector

Method build_index()

Builds the search index.

Usage
NgtIndex$build_index(num_threads = 1L, target_size_of_graph = 0L)
Arguments
num_threads

number of threads to be used for building a search index.

target_size_of_graph

target size of the graph.

Method batch_insert()

Inserts data into the index and build a search index.

Usage
NgtIndex$batch_insert(data, num_threads = 1L)
Arguments
data

data to be inserted.

num_threads

number of threads to be used for building the index.

Returns

the IDs of the inserted objects are returned invisibly.

Method insert()

Inserts a vector to the index. To search with the index, you need to call build_index after call this method.

Usage
NgtIndex$insert(vec)
Arguments
vec

vector to be inserted.

Returns

the ID of the inserted object is returned invisibly.

Method export_index()

Exports an index to a file.

Usage
NgtIndex$export_index(path)
Arguments
path

path to save the index.

Method import_index()

Imports an index from a file.

Usage
NgtIndex$import_index(path)
Arguments
path

path to load the index.

Method close()

Closes the index.

Usage
NgtIndex$close()