File IndexHNSW.h

namespace faiss

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree.

Throughout the library, vectors are provided as float * pointers. Most algorithms can be optimized when several vectors are processed (added/searched) together in a batch. In this case, they are passed in as a matrix. When n vectors of size d are provided as float * x, component j of vector i is

x[ i * d + j ]

where 0 <= i < n and 0 <= j < d. In other words, matrices are always compact. When specifying the size of the matrix, we call it an n*d matrix, which implies a row-major storage.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. I/O functions can read/write to a filename, a file handle or to an object that abstracts the medium.

The read functions return objects that should be deallocated with delete. All references within these objectes are owned by the object.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. Definition of inverted lists + a few common classes that implement the interface.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. Since IVF (inverted file) indexes are of so much use for large-scale use cases, we group a few functions related to them in this small library. Most functions work both on IndexIVFs and IndexIVFs embedded within an IndexPreTransform.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. In this file are the implementations of extra metrics beyond L2 and inner product

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. Defines a few objects that apply transformations to a set of vectors Often these are pre-processing steps.

struct IndexHNSW : public faiss::Index 

#include <IndexHNSW.h>

The HNSW index is a normal random-access index with a HNSW link structure built on top

Subclassed by faiss::IndexHNSW2Level, faiss::IndexHNSWFlat, faiss::IndexHNSWPQ, faiss::IndexHNSWSQ

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

explicit IndexHNSW(int d = 0, int M = 32, MetricType metric = METRIC_L2)

explicit IndexHNSW(Index *storage, int M = 32)

~IndexHNSW() override

virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters: x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override: Trains the storage if needed.

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels) const override: entry point for search

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters

key – id of the vector to reconstruct
recons – reconstucted vector (size d)

virtual void reset() override: removes all elements from the database.

void shrink_level_0_neighbors(int size)

void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1) const

Perform search only on level 0, given the starting points for each vertex.

Parameters: search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I): alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests): alternative graph building

void reorder_links()

void link_singletons()

Public Members

HNSW hnsw

bool own_fields

Index *storage

ReconstructFromNeighbors *reconstruct_from_neighbors

struct IndexHNSW2Level : public faiss::IndexHNSW 

#include <IndexHNSW.h>

2-level code structure with fast random access

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSW2Level()

IndexHNSW2Level(Index *quantizer, size_t nlist, int m_pq, int M)

void flip_to_ivf()

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels) const override: entry point for search

virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters: x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override: Trains the storage if needed.

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters

key – id of the vector to reconstruct
recons – reconstucted vector (size d)

virtual void reset() override: removes all elements from the database.

void shrink_level_0_neighbors(int size)

void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1) const

Perform search only on level 0, given the starting points for each vertex.

Parameters: search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I): alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests): alternative graph building

void reorder_links()

void link_singletons()

Public Members

HNSW hnsw

bool own_fields

Index *storage

ReconstructFromNeighbors *reconstruct_from_neighbors

struct IndexHNSWFlat : public faiss::IndexHNSW 

#include <IndexHNSW.h>

Flat index topped with with a HNSW structure to access elements more efficiently.

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSWFlat()

IndexHNSWFlat(int d, int M, MetricType metric = METRIC_L2)

virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters: x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override: Trains the storage if needed.

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels) const override: entry point for search

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters

key – id of the vector to reconstruct
recons – reconstucted vector (size d)

virtual void reset() override: removes all elements from the database.

void shrink_level_0_neighbors(int size)

void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1) const

Perform search only on level 0, given the starting points for each vertex.

Parameters: search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I): alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests): alternative graph building

void reorder_links()

void link_singletons()

Public Members

HNSW hnsw

bool own_fields

Index *storage

ReconstructFromNeighbors *reconstruct_from_neighbors

struct IndexHNSWPQ : public faiss::IndexHNSW 

#include <IndexHNSW.h>

PQ index topped with with a HNSW structure to access elements more efficiently.

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSWPQ()

IndexHNSWPQ(int d, int pq_m, int M)

virtual void train(idx_t n, const float *x) override: Trains the storage if needed.

virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters: x – input matrix, size n * d

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels) const override: entry point for search

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters

key – id of the vector to reconstruct
recons – reconstucted vector (size d)

virtual void reset() override: removes all elements from the database.

void shrink_level_0_neighbors(int size)

void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1) const

Perform search only on level 0, given the starting points for each vertex.

Parameters: search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I): alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests): alternative graph building

void reorder_links()

void link_singletons()

Public Members

HNSW hnsw

bool own_fields

Index *storage

ReconstructFromNeighbors *reconstruct_from_neighbors

struct IndexHNSWSQ : public faiss::IndexHNSW 

#include <IndexHNSW.h>

SQ index topped with with a HNSW structure to access elements more efficiently.

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSWSQ()

IndexHNSWSQ(int d, ScalarQuantizer::QuantizerType qtype, int M, MetricType metric = METRIC_L2)

virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters: x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override: Trains the storage if needed.

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels) const override: entry point for search

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters

key – id of the vector to reconstruct
recons – reconstucted vector (size d)

virtual void reset() override: removes all elements from the database.

void shrink_level_0_neighbors(int size)

void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1) const

Perform search only on level 0, given the starting points for each vertex.

Parameters: search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I): alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests): alternative graph building

void reorder_links()

void link_singletons()

Public Members

HNSW hnsw

bool own_fields

Index *storage

ReconstructFromNeighbors *reconstruct_from_neighbors

struct ReconstructFromNeighbors

Public Types

typedef Index::idx_t idx_t

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

explicit ReconstructFromNeighbors(const IndexHNSW &index, size_t k = 256, size_t nsq = 1)

void add_codes(size_t n, const float *x): codes must be added in the correct order and the IndexHNSW must be populated and sorted

size_t compute_distances(size_t n, const idx_t *shortlist, const float *query, float *distances) const

void estimate_code(const float *x, storage_idx_t i, uint8_t *code) const: called by add_codes

void reconstruct(storage_idx_t i, float *x, float *tmp) const: called by compute_distances

void reconstruct_n(storage_idx_t n0, storage_idx_t ni, float *x) const

void get_neighbor_table(storage_idx_t i, float *out) const: get the M+1 -by-d table for neighbor coordinates for vector i

Public Members

const IndexHNSW &index

size_t M

size_t k

size_t nsq

size_t code_size

int k_reorder

std::vector<float> codebook

std::vector<uint8_t> codes

size_t ntotal

size_t d

size_t dsub