Pith. sign in

REVIEW 5 major objections 5 minor 47 references

Billion-scale Similarity Search Using a Hybrid Indexing Approach with Advanced Filtering

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that appending discrete filter attributes directly to dense embeddings and indexing the resulting hybrid vectors with a disk-based IVF-Flat structure lets a single CPU server perform filtered nearest-neighbor search over…

desk verdict A disk-based IVF-Flat post-filtering report whose 1.428 s timing is uninterpretable without recall; the 'hybrid vector' is never actually used in search. read the letter →

arxiv 2501.13442 v1 pith:H7WY3YHD submitted 2025-01-23 cs.IR cs.DBcs.DCcs.LG

Simeon Emanuilov , Aleksandar Dimov This is my paper · ORCID
classification cs.IRcs.DBcs.DCcs.LG
keywords similaritysearchk-nearestneighborsbillion-scalehybridvectorsfilteredIVF-FlatCPUinferenceLAION-5B
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that filtered approximate nearest-neighbor search over billion-vector datasets can be done on a single CPU server, without GPUs, by extending the standard inverted-file index instead of inventing a new data structure. The central device is the hybrid vector: each point's dense embedding is concatenated with a fixed-size vector of discrete filter attributes, and the combined vector is stored in a disk-resident IVF-Flat index. At query time, the system finds the T nearest centroids by embedding similarity, applies the filter conditions inside those inverted lists, and then computes exact distances only for the survivors. In the case study on a one-billion-vector subset of LAION image embeddings with ten synthetic attributes, the reported end-to-end search time is about 1.428 seconds on one CPU with 12 BLAS threads. The paper does not report recall, so the contribution is a speed-and-scalability claim with the retrieval-quality question left open.

What carries the argument

The load-bearing object is the hybrid vector $h_i=[x_i\,\|\,a_i]$, the concatenation of a $D$-dimensional core embedding with an $M$-dimensional attribute vector, stored in a disk-resident IVF-Flat index, an inverted file whose Voronoi cells each hold a flat list of full vectors. Concatenation gives one record one index entry, so filtering metadata rides along with the embedding and no second index is needed. Search is carried by the $T$-nearest-centroid step (here $T=7$), which prunes the space before filtering, and by dynamic loading, which brings into RAM only the filtered candidates from the selected lists. The filtering step itself consumes most of the reported query time, 1.09 of 1.428 seconds, so the bottleneck is attribute comparison rather than vector distance computation.

What would settle it

Run the method on a billion-scale sample, choose a filter value that is common only outside the seven nearest cells, compare the returned top-k with an exhaustive filtered scan, and compute recall@k; if recall collapses, the 1.4-second claim is a speed-only result without a quality guarantee.

Watch

Extended reading notes

Core claim

The discovery, on the paper's own terms, is that a classical IVF-Flat index can carry multi-dimensional filters almost for free by storing hybrid vectors $h_i=[x_i\,\|\,a_i]$ and by filtering inside the $T$ nearest inverted lists before any distance computation. The search pipeline is explicit: select $T$ nearest centroids using the embedding part of the query, apply the attribute conditions through an in-memory filter structure, compute BLAS-accelerated exact distances for the surviving vectors in those lists, and merge the per-list top-k results. With K = 32,000 centroids, T = 7, M = 10 synthetic attributes, and 768-dimensional CLIP embeddings, the one-billion-vector case study reports 0.008 seconds for centroid search, 1.090 seconds for filtering, 0.330 seconds for detailed cluster search, and 1.428 seconds total. The paper also reports that mainstream PostgreSQL-based vector extensions could not build a 15-million-vector 768-dimensional index on the same hardware, which frames the method as filling a scalability gap.

Load-bearing premise

Setting T=7 assumes that every vector satisfying the filter is concentrated in the seven Voronoi cells whose centroids are closest to the query embedding; if matching vectors live elsewhere, the scan never sees them, and the paper reports no recall measurement to bound that risk.

Editorial extensions

If this is right

  • Filtered k-nearest-neighbor search over a billion vectors becomes a CPU-only, disk-resident operation needing no GPU for inference.
  • Query time on the case-study configuration is dominated by filtering (1.090 s) rather than distance computation (0.330 s), so speeding up attribute comparisons is the most direct route to faster end-to-end search.
  • Adding a vector is local: assign the hybrid vector to its nearest centroid and append it to that flat list, so incremental updates do not require rebuilding the entire billion-scale index.
  • If the same attribute encoding is extended to interval-tree range conditions and multi-attribute logical expressions, the structure is positioned to answer SQL-like filtered queries; the case study only demonstrates exact-match attributes.
  • Because the index is disk-based and loads only selected lists, the data can exceed available RAM, which is the practical precondition for billion-scale CPU deployment.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The absence of a recall measurement means the practical ceiling is set by filter selectivity: the rarer the matching attribute values, the less likely the $T=7$ nearest Voronoi cells contain them, and recall should degrade precisely on the most selective queries.
  • The index partitions by core embeddings only, so the filter distribution across cells is not controlled at construction time; a natural follow-up would be to bias list assignment or centroid selection by filter co-occurrence to protect rare-attribute recall.
  • The reported 1.428 s is search-only; index construction took hours on CPU even with MiniBatchKMeans, so total deployment cost includes an expensive offline build that the paper treats as separate from query cost.
  • A direct test of the claimed generality is to run range and multi-attribute logical filters with varying $T$ and measure recall against exhaustive filtered search; the paper states these filters are supported but demonstrates only exact matches.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes a CPU-based, disk-based extension of IVF-Flat intended to support billion-scale similarity search with multi-dimensional filtering. The method constructs "hybrid vectors" by concatenating dense embeddings with discrete attribute vectors, clusters core vectors into inverted lists, then at query time selects T nearest centroids by embedding similarity, filters the candidates inside those lists, and computes exact distances. The empirical part is a case study on a 1-billion-vector LAION subset with synthetic attributes, reporting a total search time of about 1.428 s after BLAS parallelization, with no accuracy measurement.

Significance. If the reported result were fully validated, it would be practically relevant: a single CPU server answering filtered approximate nearest-neighbor queries over one billion vectors in about 1.4 seconds is an attractive cost profile. The paper is clearly written, addresses a real problem, and provides useful implementation details such as parameter values and hardware configuration. However, the evidence presented is far too thin to support the abstract's claims of effectiveness and complex filtering: there is no recall or precision measurement, no ground-truth definition, no query workload, no baseline comparison, and no analysis of the central pruning assumption. The paper also gives explicit credit to the LAION pre-existing kNN index in Section 5.2, which makes it unclear whether the proposed index construction was actually evaluated. As it stands, the manuscript is a plausible engineering description rather than a validated research claim.

major comments (5)
  1. [Section 5.3, Table 2] The only reported search result is a latency breakdown totaling 1.428 s. There is no recall@k, precision, or any other accuracy metric, no description of the query set, no ground truth, no baselines, and no repeated-run statistics. Without an accuracy measurement, the headline latency is uninterpretable: it could reflect fast retrieval of mostly irrelevant items. This directly undermines the abstract's claim of demonstrating effectiveness.
  2. [Section 4.4, Steps 2 and 3] The search procedure prunes to the T=7 nearest centroids using only the core query vector, and applies the filter only inside those selected lists. Correctness therefore depends on the unstated assumption that filter-satisfying nearest neighbors are concentrated in the top-T embedding-similar Voronoi cells. Section 5.1 generates synthetic attributes uniformly and independently of the CLIP embeddings, so filter matches are geometrically uncorrelated with embedding neighborhoods; for selective filters they could be scattered across many cells, causing recall to collapse. The paper provides no recall-versus-T experiment, no bound, and no other evidence to support this assumption.
  3. [Section 5.2] The case study states that the authors "utilized the pre-existing kNN index provided with the LAION-5B dataset and a few processing steps (e.g., merging with filters)". This leaves it unclear whether the evaluated index was constructed by the proposed method (K-means on core vectors with hybrid vectors and flat lists) or is an external FAISS index. If the latter, the experiment does not validate the proposed hybrid index construction at all. The paper needs to clarify exactly which index was used and, if necessary, build the proposed index from scratch.
  4. [Sections 4.1 and 4.4] The hybrid vector h_i = [x_i || a_i] is never used for distance computation or for centroid selection: Step 2 of the search selects centroids by the distance on the x_input part, Step 4 computes distances between x_input and the core vectors, and Step 3 applies filters via the separate attribute vector. The method therefore reduces to post-filtering IVF-Flat with a metadata sidecar, and the claimed novelty of a hybrid representation is not substantiated. The authors should either use the hybrid vector in partition selection and distance computation or explicitly frame the contribution as post-filtering on an IVF-Flat index.
  5. [Section 5.4] The limitations section admits that "we primarily tested with an exact match for the attribute vector" and that the attributes are synthetic random integers independent of the embeddings. Yet the abstract and Section 3.4 promise support for complex, multi-dimensional, SQL-like filters, range queries via interval trees, and multi-attribute logical operations. None of these claimed capabilities is evaluated. The empirical evidence covers only equality filtering on synthetic attributes, which is the easiest case for the proposed pruning strategy.
minor comments (5)
  1. [Section 2.3] The claim that pgvector and pgvectorscale became unresponsive on a 15-million-vector sample is supported by reference [29], which is the first author's own blog post. This should either be substantiated by reproducible experiments or removed, since it is not independent evidence.
  2. [Section 4.3 vs Section 4.4] Section 4.3 describes disk-based storage and intelligent caching as a "future direction", while Section 4.4 presents a disk-based, dynamically loaded structure as part of the current method. These statements should be reconciled.
  3. [Section 4.4, Step 2] The text says the T nearest centroids are identified for the "hybrid query vector" q_h, but then specifies that the distance is based only on the x_input part. This wording is confusing and should be rewritten to say that centroid selection uses only the core vector.
  4. [Section 5, Table 1] Table 1 lists the parameters but does not explain how T=7 was chosen or what trade-off was considered. A sentence explaining the choice would help, especially since T is the main accuracy-critical parameter.
  5. [General] The paper would benefit from a comparison with at least one standard baseline, such as IVF-Flat without filtering or Filtered-DiskANN, even on a small labeled subset, to calibrate the reported latency.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity: the hybrid-index search is a directly measured IVF-Flat post-filtering pipeline; the only circularity-adjacent item is a minor, non-load-bearing self-citation.

  1. other [Section 2.3, reference [29]]
    "These attempts were unsuccessful on CPU-based hardware, resulting in system unresponsiveness and excessive processing times [29]."

    The support for the preliminary claim that pgvector and pgvectorscale cannot handle the 15M-vector sample is reference [29], a blog post by the first author. This is a self-citation, but it is not load-bearing for the paper's central result: the proposed method's 1.428 s search time and its search procedure are measured and specified in Sections 4 and 5, not derived from the cited post. The citation supports only motivational context, so it does not make the derivation circular.

full rationale

The paper contains no fitted parameter that is later called a prediction, no quantity defined in terms of the quantity it is supposed to derive, and no uniqueness theorem imported from the authors' prior work. The search algorithm (Section 4.4) is a standard IVF-Flat candidate-generation step followed by attribute filtering and exact distance computation; the hybrid vector is constructed by concatenation, and the paper's own equations show that centroid selection uses only the core-vector part. The case-study timing (Table 2) is a direct measurement. The only circularity-adjacent issue is reference [29], a self-published post by the first author used to support the preliminary claim that pgvector/pgvectorscale fail on a 15M-vector sample; this is motivational rather than load-bearing, because the main evaluation is self-contained in Section 5. Missing recall measurements and the untested T=7 pruning assumption are correctness and validity concerns, not circularity.

Assumptions & free parameters 2 free parameters · 3 assumptions · 1 invented entities

The central design rests on standard IVF heuristics plus three unverified domain assumptions. The hybrid vector is definitional and unused in scoring, and the method's effectiveness depends on an assumption about filter distribution across Voronoi cells that is not tested.

free parameters (2)
  • T (number of nearest centroids scanned) = 7
    Set by hand in Section 5 as a balance between accuracy and efficiency; no ablation or recall measurement justifies the choice.
  • K (number of centroids) = 32,000
    Set to approximately sqrt(N) following a common heuristic in Section 4.2. The paper does not tune K to the data or measure its effect on filtered recall.
assumptions (3)
  • domain assumption Vectors satisfying the filter are present in the T nearest centroids to the query
    Section 4.4 Steps 2 and 3 apply filters only after centroid selection. If filtered vectors sit outside the scanned cells, recall collapses. This is never measured.
  • domain assumption Fixed-size integer encoding preserves the expressiveness of arbitrary filter attributes
    Section 3.4 claims one-hot encoding and adaptive binning support exact and range queries, but only exact match on synthetic random integers is tested.
  • ad hoc to paper The precomputed LAION kNN index is a valid stand-in for the proposed construction
    Section 5.2 uses the dataset's existing FAISS index rather than building the hybrid index from scratch, so the reported search budget may not reflect the proposed construction pipeline.
invented entities (1)
  • Hybrid vector h_i = [x_i || a_i]
    purpose: Claimed unified representation integrating embeddings and filter attributes into one index structure
    Defined in Section 4.1 but never used in centroid selection or distance scoring (Section 4.4 Steps 2 and 4). It functions as a storage association rather than an independently validated construct.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Billion-scale Similarity Search Using a Hybrid Indexing Approach with Advanced Filtering." pith.science (2026). https://pith.science/paper/H7WY3YHD

@misc{pith2026250113442,
  author       = {Pith},
  title        = {Pith review of: Billion-scale Similarity Search Using a Hybrid Indexing Approach with Advanced Filtering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H7WY3YHD}},
  note         = {Machine review of arXiv:2501.13442}
}
read the original abstract

This paper presents a novel approach for similarity search with complex filtering capabilities on billion-scale datasets, optimized for CPU inference. Our method extends the classical IVF-Flat index structure to integrate multi-dimensional filters. The proposed algorithm combines dense embeddings with discrete filtering attributes, enabling fast retrieval in high-dimensional spaces. Designed specifically for CPU-based systems, our disk-based approach offers a cost-effective solution for large-scale similarity search. We demonstrate the effectiveness of our method through a case study, showcasing its potential for various practical uses.

Figures

Figures reproduced from arXiv: 2501.13442 by the authors.

Figure 1
Figure 1. How the hybrid vectors are constructed In [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Centroids, index elements, filters The number of centroids (𝐾) in the IVF-Flat index affects the trade-off between index size, construction time, and search efficiency [31]. A common heuristic is to set 𝐾 to 𝑁/1000 for datasets up to 1 million vectors, or sqrt(𝑁) for larger datasets, where 𝑁 is the total number of vectors (Section 4.3) [27]. 4.3. Centroid selection and search parameters The number of centroids (𝐾) i… view at source ↗
Figure 3
Figure 3. Searching procedure One key adaptation in our approach is the dynamic, memory-efficient loading strategy. During a search operation, only the vectors from the 𝑇 most relevant inverted lists that meet the filtering conditions are loaded into memory. This selective loading significantly reduces memory usage while maintaining high performance. Our disk-based, dynamically loaded structure effectively manages billion-sca… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 42 canonical work pages

  1. [29]

    S i n g h, A., S. J. S u b r a m a n ya, R. K r i s h n a s w a my, H. V. S i m h a d ri. FreshDiskANN: A Fast and Accurate Graph -Based ANN Index for Streaming Similarity Search. arXiv Preprint arXiv:2105.09613, 2021

    arXiv 2021

  2. [1]

    As datasets grow to billions of entries, the challenge of performing efficient searches on high -dimensional vectors becomes increasingly complex [4]

    Introduction Similarity search, the task of finding similar vectors, has become a fundamental operation in machine learning, with applications in recommendation engines, semantic search systems, and more [1-3]. As datasets grow to billions of entries, the challenge of performing efficient searches on high -dimensional vectors becomes increasingly complex ...

    work page

  3. [2]

    This section reviews the most relevant literature to our work, focusing on techniques for ANN search and filtering in billion-scale datasets

    Related work Similarity search has witnessed significant advancements in recent years, driven by the increasing prevalence of high -dimensional data in various domains [9]. This section reviews the most relevant literature to our work, focusing on techniques for ANN search and filtering in billion-scale datasets. 2.1. Recent advancements in similarity sea...

    work page

  4. [3]

    Background Before exploring the details of our proposed approach, we provide a brief overview of the important notations used throughout this paper. 3.1. Inverted File index (IVF) IVF is a fundamental approach for ANN search in high-dimensional spaces [6]. IVF partitions the search space into Voronoi cells ( Section 3.3), each associated with a centroid v...

    work page

  5. [4]

    A set of 𝐾 centroids, denoted as 𝐶 = {𝑐1, 𝑐2, … , 𝑐𝐾}, where each centroid 𝑐𝑘  𝐷 represents the center of the k-th Voronoi cell, k=1, …, K, and 𝐷 represents the dimensionality of the vectors in the dataset

    work page

  6. [5]

    vector” and “embedding

    A set of 𝐾 inverted lists, denoted as 𝐿 = {𝐿1, 𝐿2, … , 𝐿𝐾}, where each list 𝐿𝑘 contains the identifiers of the vectors assigned to the k-th centroid. During indexing, the identifiers (pointers) of dataset vectors are assigned to their nearest centroids, forming these inverted lists (Section 3.3). At query time, the search is limited to a subset of the mos...

    work page

  7. [6]

    Proposed approach In this section, we present the proposed approach for a cost -efficient, large -scale similarity search with complex filtering capabilities. 4.1. Constructing hybrid vectors First, we need to construct the hybrid vector, mentioned in Section 3.5. The hybrid vectors are denoted as 𝐻 = {ℎ1, ℎ2, … , ℎ𝑁}, where each vector ℎ𝑖  (𝐷+𝑀), i = 1...

    work page

  8. [7]

    K-Means or MiniBatchKMeans [30] clustering is performed on the core vectors 𝑥𝑖 ∈  𝐷 to obtain 𝐾 cluster centroids, denoted as 𝐶 = {𝑐1, 𝑐2, … , 𝑐𝐾}, where each centroid 𝑐𝑘  𝐷

    Centroid computation. K-Means or MiniBatchKMeans [30] clustering is performed on the core vectors 𝑥𝑖 ∈  𝐷 to obtain 𝐾 cluster centroids, denoted as 𝐶 = {𝑐1, 𝑐2, … , 𝑐𝐾}, where each centroid 𝑐𝑘  𝐷. These centroids serve as the representatives of the inverted lists and are used to guide the search process

    work page

Show all 47 references
  1. [8]

    Each core vector 𝑥𝑖 is assigned to its nearest centroid cₖ based on a distance metric 𝑑(·,·), e.g., cosine similarity

    Vector assignment. Each core vector 𝑥𝑖 is assigned to its nearest centroid cₖ based on a distance metric 𝑑(·,·), e.g., cosine similarity. This forms the inverted lists, denoted as 𝐿 = {𝐿1, 𝐿2, … , 𝐿𝐾}, where each list Lₖ, k = 1, …, K, contains the indices of the core vectors a...

  2. [9]

    For each inverted list 𝐿𝑘, the full core vectors are stored

    Flat index construction. For each inverted list 𝐿𝑘, the full core vectors are stored. This flat storage approach means that the complete vector data is retained, as opposed to pointers or compressed, quantized representations

  3. [10]

    In addition to the core vectors, the index structure maintains the corresponding filter attributes for each vector

    Filter attribute association . In addition to the core vectors, the index structure maintains the corresponding filter attributes for each vector. These attributes are stored in a manner that preserves their association with the core vectors, allowing for efficient filtering o...

  4. [11]

    Building upon these initial findings, we conducted a comprehensive case study to demonstrate the effectiveness and efficiency of our proposed algorithm

    Case study As stated in Section 2.3 , our initial experiments demonstrated the limitations of existing methods like pgvector and pgvectorscale when dealing with large -scale datasets. Building upon these initial findings, we conducted a comprehensive case study to demonstrate ...

  5. [12]

    Our approach extends the classical IVF-Flat structure by introducing hybrid vectors that integrate dense embeddings and discrete filtering attributes

    Conclusion In this paper, we presented a novel algorithm for cost-efficient similarity search with complex filtering capabilities on billion-scale datasets, optimized for CPU inference. Our approach extends the classical IVF-Flat structure by introducing hybrid vectors that in...

  6. [13]

    Similarity Search and Data Mining: Database Techniques Supporting Next Decade ’s Applications

    B ö hm, C. Similarity Search and Data Mining: Database Techniques Supporting Next Decade ’s Applications. – Unit for Database Systems, University for Health Informatics and Technology

  7. [14]

    S e v i l m is, B

    F o s t e r, C., B. S e v i l m is, B. K i m i a. Generalized Relative Neighborhood Graph (GRNG) for Similarity Search. – In: Proc. of International Conference on Similarity Search and Applications, Cham: Springer International Publishing, September 2022, pp. 133-149

    2022

  8. [15]

    Auto-Tuning Similarity Search Algorithms on Multi-Core Architectures

    G e d i k, B. Auto-Tuning Similarity Search Algorithms on Multi-Core Architectures. – International Journal of Parallel Programming, Vol. 41, 2013, No 5, pp. 595-620

    2013

  9. [16]

    K h o r s h i di, M. S., N. Y a z d a n j ue, H. G h a r o un, D. Y a z d a ni, M. R. N i k o o, F. C h e n, A. H. G a n d o mi. Semantic-Preserving Feature Partitioning for Multi -View Ensemble Learning. arXiv Preprint arXiv:2401.06251, 2024

    2024 arXiv

  10. [17]

    The Curse of Dimensionality

    K ö p p en, M. The Curse of Dimensionality. – In: Proc. of 5th Online World Conference on Soft Computing in Industrial Applications (WSC5’00), Vol. 1, September 2000, pp. 4-8

    2000

  11. [18]

    D o u z e, C

    J é g ou, H., M. D o u z e, C. S c h m id. Product Quantization for Nearest Neighbor Search. – IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 33, 2011, No 1, pp. 117-128. https://doi.org/10.1109/TPAMI.2010.57

    2011 doi

  12. [19]

    M a l k ov, Y. A., D. A. Y a s h u n in. Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs. – IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 42, 2020, No 4, pp. 824-836. https://doi.org/10.1109/TPAMI....

    2020

  13. [20]

    Voronoi Cells

    S t u r m f e ls, B. Voronoi Cells . University of California, Berkeley , 2023 (Last accessed: 26.07.2024). https://math.berkeley.edu/~bernd/wednesday.pdf 57

    2023

  14. [21]

    Z o u m p a t i a n os, T

    E c h i h a bi, K., K. Z o u m p a t i a n os, T. P a l p a n as. High-Dimensional Similarity Search for Scalable Data Science. – In: Proc. of 37th IEEE International Conference on Data Engineering (ICDE’21), April 2021, pp. 2369-2372

    2021

  15. [22]

    B a b e n ko, Y

    B a r a n c h uk, D., A. B a b e n ko, Y. M a l k ov. Revisiting the Inverted Indices for Billion-Scale Approximate Nearest Neighbors . – In: Proc. of European Conference on Computer Vision (ECCV’18), 2018, pp. 202-216

    2018

  16. [23]

    D o u z e, H

    J o h n s on, J., M. D o u z e, H. J é g ou. Billion-Scale Similarity Search with GPUs. – IEEE Transactions on Big Data, Vol. 7, 2019, No 3, pp. 535-547

    2019

  17. [24]

    W a n g, M

    Y a n g, K., H. W a n g, M. D u, Z. W a n g, Z. Ta n, Y. X i a o. Hierarchical Link and Code: Efficient Similarity Search for Billion-Scale Image Sets. – In: PG (Short Papers, Posters, and Work-in- Progress Papers), 2021, pp. 81-86

    2021

  18. [25]

    Z h a n g, P., Z. Li u, S. X i a o, Z. Do u, J. Ya o. Hybrid Inverted Index Is a Robust Accelerator for Dense Retrieval. – In: Proc. of 2023 Conference on Empirical Methods in Natural Language Processing, December 2023, pp. 1877-1888

    2023

  19. [26]

    K a r i a, V

    G o l l a p u di, S., N. K a r i a, V. S i v a s h a n k ar, R. K r i s h n a s w a my, N. B e g w a ni, S. Ra z, Y. Li n, Y. Z h a n g, N. M a h a p a t ro, P. S r i n i v a s an, A. S i n g h. Filtered- DiskANN: Graph Algorithms for Approximate Nearest Neighbor Search with F...

    2023

  20. [27]

    D e v v r it, H

    J a y a r a m S u b r a m a n y a, S., F. D e v v r it, H. V. S i m h a d ri, R. K r i s h n a w a my, R. K a d e k o di. DiskANN: Fast Accurate Billion-Point Nearest Neighbor Search on a Single Node. – Advances in Neural Information Processing Systems, 2019, 32

    2019

  21. [28]

    A l-D u j a i li, M. J., H. J a b a r S a b a t A h i l y. A New Hybrid Model to Predict Human Age Estimation from Face Images Based on Supervised Machine Learning Algorithms. – Cybernetics and Information Technologies, Vol. 23, 2023, No 2, pp. 20-33

    2023

  22. [30]

    T u r m u k h a m e t o va, N

    S u n d a r am, N., A. T u r m u k h a m e t o va, N. S a t i sh, T. M o s t ak, P. I n d y k, S. M a d d en, P. D u b e y. Streaming Similarity Search over One Billion Tweets Using Parallel Locality-Sensitive Hashing. – In: Proc. of VLDB Endowment, Vol. 6, 2013, No 14, pp. 1930-1941

    2013

  23. [31]

    5mins on Postgres: Vectors with PGvector

    Pganalyze Blog. 5mins on Postgres: Vectors with PGvector. 2021 (Last accessed: 26.07.2024). https://pganalyze.com/blog/5mins-postgres-vectors-pgvector

    2021

  24. [32]

    AnalyticDB: Real-time OLAP Database

    Alibaba Cloud. AnalyticDB: Real-time OLAP Database. 2021 (Last accessed: 26.07.2024). https://www.alibabacloud.com/product/analyticdb

    2021

  25. [33]

    TOAST (Last accessed: 26.07.2024)

    PostgreSQL Documentation: 16: 73.2. TOAST (Last accessed: 26.07.2024). https://www.postgresql.org/docs/current/storage-toast.html

    2024

  26. [34]

    Accelerating Vector Search Using GPU-Powered Indexes with RAPIDS Raft

    NVIDIA Developer Blog. Accelerating Vector Search Using GPU-Powered Indexes with RAPIDS Raft. 2021 (Last accessed: 26.07.2024). https://developer.nvidia.com/blog/accelerating-vector-search-using-gpu-powered- indexes-with-rapids-raft/

    2021

  27. [35]

    Accelerated Vector Search: Approximating with RAPIDS raft IVF -Flat

    NVIDIA Developer Blog. Accelerated Vector Search: Approximating with RAPIDS raft IVF -Flat. 2021 (Last accessed: 26.07.2024). https://developer.nvidia.com/blog/accelerated-vector-search-approximating-with- rapids-raft-ivf-flat/

    2021

  28. [36]

    D o s h i, I., D. Da s, A. B h u t a ni, R. K u m a r, R. B h a t t, N. B a l a s u b r a m a n i an. LANNS: A Web-Scale Approximate Nearest Neighbor Lookup System. – Proc. of VLDB Endowment, Vol. 15, 2021, pp. 850-858

    2021

  29. [37]

    H ö g q v i st

    S c h ä f er, P., M. H ö g q v i st. SFA: A Symbolic Fourier Approximation and Index for Similarity Search in High Dimensional Datasets . – In: Proc. of 15th International Conference on Extending Database Technology, March 2012, pp. 516-527

    2012

  30. [38]

    G u z h va, C

    D o u z e, M., A. G u z h va, C. D e n g, J. J o h n s on, G. S z i l v a sy, P. E. M a z a r é, M. L o m e li, L. H o s s e i ni, H. J é g ou. The FAISS Library. arXiv Preprint arXiv:2401.08281, 2024

    2024 arXiv

  31. [39]

    https://github.com/pgvector/pgvector 58

    Pgvector (Last accessed: 26.07.2024). https://github.com/pgvector/pgvector 58

    2024

  32. [40]

    https://github.com/timescale/pgvectorscale

    Pgvectorscale (Last accessed: 16.10.2024). https://github.com/timescale/pgvectorscale

    2024

  33. [41]

    pgvectorscale – Accelerating AI Development with High-Performance Vector Search (Last accessed: 16.10.2024)

    E m a n u i l ov, S. pgvectorscale – Accelerating AI Development with High-Performance Vector Search (Last accessed: 16.10.2024). https://unfoldai.com/pgvectorscale-extension-for-ai-apps/

    2024

  34. [42]

    https://scikit- learn.org/stable/modules/generated/sklearn.cluster.MiniBatchKMeans.html

    Scikit Learn, MiniBatchKMeans API Reference and Documentation (Last accessed: 26.07.2024). https://scikit- learn.org/stable/modules/generated/sklearn.cluster.MiniBatchKMeans.html

    2024

  35. [43]

    M e d i ni, A

    G u p t a, G., T. M e d i ni, A. S h r i v a s t a va, A. J. S m o l a. Bliss: A Billion Scale Index Using Iterative Re -Partitioning. – In: Proc. of 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, August 2022, pp. 486-495

    2022

  36. [44]

    V e n c u, R

    S c h u h m a nn, C., R. V e n c u, R. B e a u m o nt, R. K a c z m a r c z yk, C. M u l l is, A. K a t t a, T. C o o m b es, J. J i t s ev, A. K o m a t s u z a ki. LAION-5B: An Open Large-Scale Dataset for Training Next-Generation Image-Text Models. arXiv Preprint arXiv:2210...

    arXiv

  37. [45]

    R a d f o rd, A., J. W. Ki m, C. H a l l a cy, A. R a m e sh, G. Go h, S. A g a r w al, G. S a s t ry, A. A s k e ll, P. M i s h k in, J. C l a r k, G. K r u e g er. Learning Transferable Visual Models from Natural Language Supervision . – In: Proc. of International Conference...

    2021

  38. [46]

    Semantic Search at Billions Scale (Last accessed: 26.07.2024)

    B e a u m o nt, R. Semantic Search at Billions Scale (Last accessed: 26.07.2024). https://rom1504.medium.com/semantic-search-at-billions-scale-95f21695689a

    2024

  39. [47]

    S e l v i t o pi, M

    A z a d, A., O. S e l v i t o pi, M. T. H u s s a in, J. R. G i l b e rt, A. B u l u ç. Combinatorial BLAS 2.0: Scaling Combinatorial Algorithms on Distributed-Memory Systems. – IEEE Transactions on Parallel and Distributed Systems, Vol. 33, 2021, No 4, pp. 989-1001. Received:...

    2021

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.