Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

DARTH: Declarative Recall Through Early Termination for Approximate Nearest Neighbor Search

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read DARTH terminates ANN searches early, per query, the moment a trained model predicts the user's target recall has been reached, cutting search time by up to 14.6x on HNSW and 41.8x on IVF while meeting the declared recall target.

desk verdict A well-executed empirical early-termination method whose own data contradict the headline declarative-recall guarantee: ~15% of queries miss the target and no confidence bound is given. read the letter →

arxiv 2505.19001 v2 pith:ADPTDLPT submitted 2025-05-25 cs.DB

Manos Chatzakis , Yannis Papakonstantinou , Themis Palpanas This is my paper
classification cs.DB
keywords approximatenearestneighborsearchdeclarativerecallearlyterminationHNSWIVFgradientboostingpredictionqueryhardness
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

DARTH claims that approximate nearest neighbor search can honor a user-declared recall target, such as 0.90, without per-dataset parameter tuning by stopping each query at the moment its recall reaches the target. The method integrates an early termination loop into popular ANNS algorithms like HNSW and IVF, driven by a Gradient Boosting Decision Tree that predicts current recall from search-progress and nearest-neighbor-distance features. Because each query stops at its own predicted crossing point, easy queries terminate early while hard queries search longer, adapting to query hardness automatically. The authors report that on five standard datasets DARTH meets or exceeds targets such as 0.80-0.99 with average speedups of 6.8x for HNSW and 13.6x for IVF, while performing on average only 5% more distance calculations than the per-query optimal stopping point. The appeal is that applications declare the quality they need and get it without experimenting with algorithm-specific parameters.

What carries the argument

The load-bearing component is the recall predictor: a GBDT model (100 trees, trained with LightGBM in about 1-5 minutes) that takes eleven input features -- three index features ($n_{step}$, number of distance calculations, number of result-set inserts), three nearest-neighbor distance features (first, closest, and k-th/furthest NN distances), and five NN distance statistics (average, variance, median, 25th and 75th percentiles) -- and outputs an estimate $\hat{R}$ of the current recall. The other key mechanism is the adaptive prediction interval formula $pi = mpi + (ipi-mpi)\cdot(R_t - \hat{R})$, which shrinks the interval between predictor calls as the predicted recall nears the target, balancing prediction overhead against termination accuracy.

What would settle it

Run DARTH with a target recall of 0.90 on a held-out query workload that includes 20% Gaussian-noise-corrupted queries and compute the fraction of queries whose true recall falls below 0.90 (RQUT). If that fraction exceeds a user-specified threshold (e.g., 5%) for a workload where the plain index can reach 0.99 recall on every query, the early-termination guarantee would be violated.

Watch

Extended reading notes

Core claim

DARTH establishes that declarative recall in ANNS can be achieved at run time by early termination, rather than by offline tuning of index or search parameters. Its central claim is that a small set of search features -- the search step count, number of distance calculations, insertion count, and statistics of the distances of the nearest neighbors found so far -- suffice to train a GBDT that accurately estimates a query's current recall at any point during the base-layer search. The search stops as soon as the predicted recall reaches the target $R_t$, with the prediction invoked at adaptively chosen intervals so that the model is consulted more frequently as the predicted recall approaches the target. The authors claim this is the first early-termination method that natively supports any target recall attainable by the index, requires no hyperparameter tuning (a heuristic sets the interval parameters from training queries), and remains robust for harder and out-of-distribution query workloads, in contrast to fixed-parameter competitors.

Load-bearing premise

The recall predictor trained on 10,000 sample queries must generalize well enough to every future query -- including harder and out-of-distribution ones -- that stopping when the predicted recall reaches the target actually means the true recall has reached it.

Editorial extensions

If this is right

  • Applications can specify a recall target directly, so tuning effort shifts from per-dataset parameter search to a one-time predictor training on about 10K sample queries.
  • Search performance automatically concentrates effort on hard queries; easy queries stop early, so average latency drops while worst-query recall quality stays close to the target.
  • Because the termination decision is per-query, the method can be layered on existing HNSW and IVF implementations with only a few lines of search-loop modification, and the same feature set extends to other graph-based ANNS methods.
  • The near-optimal distance count (5% above the per-query optimal) implies that, on average, the gain from early termination is close to the theoretical best for the tested workloads.
  • The method's robustness on noisy and out-of-distribution queries suggests that declarative recall could be offered as a service-level objective in production vector databases without re-tuning per workload shift.

Reading between the lines

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

  • A natural extension the authors do not explore is using the same predictor to serve multiple recall targets simultaneously, since a single invocation supports all $R_t$ values below the predicted recall; the query could return different result sets to different consumers with different targets.
  • One could test whether the feature set transfers across datasets out-of-the-box: if a predictor trained on one dataset's queries generalizes to another index built on the same data distribution, the per-dataset training cost could be amortized further.
  • The adaptive interval formula is linear in $(R_t - \hat{R})$; an asymmetric or learned schedule might reduce the observed miss rate without increasing the number of predictor calls.
  • The 5%-above-optimal distance count is an average; for queries where the predictor is overconfident, the method stops below the target, so production deployments would likely combine DARTH with a small safety margin on the target or a fallback completion policy.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. DARTH proposes to augment HNSW and IVF searches with a GBDT-based recall predictor that is invoked at adaptively chosen intervals; the search terminates as soon as the predicted recall reaches a user-declared target R_t. The paper claims that this is the first declarative-recall solution via early termination, that it meets targets 'always' or 'with high probability,' and that it does so with speedups up to 14.6x for HNSW and 41.8x for IVF while performing about 5% more distance calculations than a per-query oracle optimum. The evaluation covers five datasets, recall targets 0.80-0.99, varying k, noisy and out-of-distribution workloads, and multiple quality measures including RQUT, RDE, NRS, P99, and Worst 1%.

Significance. If the per-query declarative guarantee were established, DARTH would be a significant contribution: it directly targets a practical pain point of ANNS parameter tuning, adapts to per-query hardness, and its speedup claims are substantial. The engineering is careful: the code is public, training-data generation takes minutes, the adaptive interval method reduces predictor calls to 6-11 per query, and the near-optimal distance-calculation claim in Figure 8 is well supported. However, as presented, the evidence supports an effective average-recall early-termination optimizer rather than a declarative per-query recall guarantee. The paper's own data show a material fraction of queries terminating below R_t, which is the load-bearing gap. The novelty claim depends on the guarantee, so resolving this gap is essential.

major comments (4)
  1. [Sections 1 and 2.3 vs. Section 4.2.3 and Figure 7] The paper's central guarantee is not delivered by the reported experiments. Section 1 states that DARTH 'is always able to meet the user-declared recall targets,' and Section 2.3 defines the objective as achieving recall at least R_t 'with high probability' without quantifying that probability. Section 4.2.3 and Figure 7 report that roughly 15% of queries terminate below R_t on the default SIFT100M workload, and Figures 13 and 20 report nonzero RQUT values for noisy and OOD workloads. These statements are internally inconsistent. The authors must either (a) define a formal coverage guarantee such as P(recall >= R_t) >= 1 - delta, calibrate the stopping rule to satisfy it, and report achieved coverage across all workloads, or (b) explicitly re-scope the contribution to average-recall optimization and remove the 'declarative' and 'always' claims.
  2. [Section 3.2.1 and Algorithm 1, lines 24-31] The stopping rule is a point-prediction threshold: DARTH terminates when the GBDT prediction R_p is at least R_t. With predictor MAE around 0.027 and R^2 around 0.88 (Table 5), errors at the decision boundary are common enough that a material fraction of queries terminate below target, as Figure 7 shows. No analysis quantifies the probability of overestimation at the boundary, and no safety margin or uncertainty estimate is used. The manuscript needs either a calibrated lower confidence bound on recall as the stopping criterion, or an explicit bound on the under-coverage probability; otherwise the connection between 'predicted recall >= R_t' and 'true recall >= R_t' is unsupported.
  3. [Section 4.2.9 and Table 5] For the T2I100M OOD workload, the predictor achieves MSE=0.029, MAE=0.079, and R^2=0.54. The text claims that DARTH 'consistently meets and surpasses all recall targets' (Figure 18a), but Figure 20 shows nonzero RQUT and even acknowledges that LAET achieves lower RQUT at R_t=0.95. With a predictor of this accuracy, the per-query guarantee is particularly implausible. Please report exact RQUT and achieved recall distributions for every R_t and k on T2I100M, and reconcile these numbers with the declarative-guarantee claim.
  4. [Section 4.2.7 and Figure 11] The robustness comparison for noisy queries is informative, but the claim that DARTH 'reaches recall very near to the declared R_t across the entire range of noise values' is only meaningful in the regime where R_t is attainable. Figure 11 shows the plain-HNSW maximum recall falling below 0.90 for noise above roughly 10-12%, so for higher noise levels the declared target cannot be met by any method. The paper should state the attainable-recall condition explicitly in the main text and report RQUT separately for the attainable and unattainable regimes, since the latter cannot by construction satisfy the declared target.
minor comments (6)
  1. [Section 4.2.2, Figure 6] The x-axis label 'GLOVE100' is inconsistent with the dataset name GLOVE1M used in Table 2 and elsewhere.
  2. [Section 4.1.1, Figure 3] The y-axis label 'MSE (10^2)' is ambiguous; the MSE values discussed in the text are around 0.003, so the plot scale and the parenthetical factor should be clarified.
  3. [Equation (1)] The adaptive interval formula should specify how non-integer values of p_i are handled, since Algorithm 1 uses 'i_dis mod p_i' and p_i is described as a number of distance calculations.
  4. [Section 3.2.2] Calling DARTH 'essentially parameter-free' is misleading: i_pi and m_pi are set from training-query statistics, and the GBDT hyperparameters (100 estimators, learning rate 0.1) are fixed choices. 'Heuristic-selected hyperparameters' would be a more accurate description.
  5. [Section 4.2.3] The sentence 'indicating that the majority of queries achieve a recall that surpasses, yet remains close to, the corresponding recall target, since roughly 15% of the queries do not meet the target' is confusing as written and should be rewritten; more importantly, the 15% figure should be reconciled with the guarantee claims made earlier in the paper.
  6. [Section 4.2.5 and Figure 9] The comparison of 'queries DARTH can answer before LAET is tuned' is not a standard quality or speed comparison; the experimental protocol should state what setup costs are included for each method, including DARTH's 10K-query training-data generation and predictor training.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: DARTH's recall predictor is a supervised empirical model validated against ground truth, and the reported below-target queries are a correctness/robustness gap, not a circular step.

full rationale

The paper's derivation chain is not circular. DARTH's central mechanism is a GBDT recall predictor trained on observations that pair search features (Table 1) with the actual recall measured at each observation time (Section 3.1.3), and the early-termination rule (Algorithm 1, lines 24-28) compares the predictor output Rp against the declared target Rt. This is an empirical estimation pipeline, not a definitional reduction: nothing in Eq. (1) or in the stopping rule forces the achieved recall to equal the predicted recall. The claimed near-optimality is validated against an independently computed per-query optimal termination point (Section 4.2.4), where the optimal is obtained by monitoring true recall after every distance calculation rather than from DARTH's own predictor. The hyperparameters ipi and mpi are fitted using training-query statistics, but they are disclosed and the comparison against the Baseline (which uses the same dists_Rt statistic as a fixed stopping point) is a fair ablation rather than a renamed version of the result. Self-citations in the related work (e.g., ProS, iSAX, and other Palpanas-group papers) are background references and are not load-bearing for the recall-prediction claim. The paper does contain an internal inconsistency: Section 2.3 promises recall at least Rt 'with high probability' and the introduction claims DARTH 'is always able to meet' targets, while Section 4.2.3 and Figure 7 report that roughly 15% of SIFT100M queries terminate below Rt, and Figure 13 shows nonzero RQUT on harder workloads. That is a substantive correctness and guarantee-formalization concern, but it is not an instance of circularity: the reported failures are empirical misses of a learned estimator, not results that are forced by the paper's own definitions or equations. Because the central derivation is self-contained and independently benchmarked, the circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central mechanism is a fitted regression model plus data-derived interval hyperparameters; no new physical entities are introduced. The key unproven premise is that the learned predictor can certify, with high probability, that the recall target is reached. The paper provides no bound on the miss rate (RQUT), and its own data show misses on 13-15% of default queries.

free parameters (4)
  • prediction interval hyperparameters ipi and mpi = dists_Rt/2 and dists_Rt/10, where dists_Rt is the average number of distance calculations needed by training queries…
    Set via a data-derived heuristic in Section 3.2.2; they control how often the predictor is invoked and are fitted per dataset and target recall.
  • training query count = 10,000
    Selected based on validation MSE plateau (Figure 3); a hand-picked design choice, not learned.
  • GBDT hyperparameters = 100 estimators, learning rate 0.1
    Fixed by hand in Section 3.1.2, not tuned per dataset.
  • IVF/T2I logging frequency = every 20/50 distance calcs (IVF), every 2 (T2I)
    Chosen to keep training data manageable in Sections 4.2.10 and 4.2.9; affects predictor resolution.
assumptions (5)
  • domain assumption The declared target recall Rt is attainable by the index for each query, i.e., Rt <= R_h_q, the recall of plain HNSW.
    Stated in Section 2.3; fails for hard/OOD queries where plain HNSW itself cannot reach Rt, as seen in Figure 11.
  • domain assumption Recall is non-decreasing during the base-layer search, so early stopping when predicted recall >= Rt implies final recall >= Rt if the prediction is perfect.
    Implicit in Algorithm 1; true because the result set only grows during search.
  • domain assumption The GBDT predictor, trained on 10K training queries from the same dataset, generalizes to unseen test queries, including harder and OOD queries.
    Standard supervised learning assumption; not proven and empirically violated for 13-15% of default queries and for high-noise workloads (Sections 4.2.3, 4.2.7).
  • domain assumption The selected input features (index progress, NN distances, NN stats) are sufficient to accurately predict current recall at any point.
    Justified only by ablation (Section 4.1.4); there is no formal analysis.
  • domain assumption Distance calculations are an adequate unit for prediction intervals and effort accounting.
    Used throughout; reasonable but not uniformly accurate across hardware.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DARTH: Declarative Recall Through Early Termination for Approximate Nearest Neighbor Search." pith.science (2026). https://pith.science/paper/ADPTDLPT

@misc{pith2026250519001,
  author       = {Pith},
  title        = {Pith review of: DARTH: Declarative Recall Through Early Termination for Approximate Nearest Neighbor Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ADPTDLPT}},
  note         = {Machine review of arXiv:2505.19001}
}
read the original abstract

Approximate Nearest Neighbor Search (ANNS) presents an inherent tradeoff between performance and recall (i.e., result quality). Each ANNS algorithm provides its own algorithm-dependent parameters to allow applications to influence the recall/performance tradeoff of their searches. This situation is doubly problematic. First, the application developers have to experiment with these algorithm-dependent parameters to fine-tune the parameters that produce the desired recall for each use case. This process usually takes a lot of effort. Even worse, the chosen parameters may produce good recall for some queries, but bad recall for hard queries. To solve these problems, we present DARTH, a method that uses target declarative recall. DARTH uses a novel method for providing target declarative recall on top of an ANNS index by employing an adaptive early termination strategy integrated into the search algorithm. Through a wide range of experiments, we demonstrate that DARTH effectively meets user-defined recall targets while achieving significant speedups, up to 14.6x (average: 6.8x; median: 5.7x) faster than the search without early termination for HNSW and up to 41.8x (average: 13.6x; median: 8.1x) for IVF. This paper appeared in ACM SIGMOD 2026.

Figures

Figures reproduced from arXiv: 2505.19001 by the authors.

Figure 1
Figure 1. Early Termination margins for target recall 0.80. Curve represents recall improvement for queries in [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. (a): Example of locating the nearest neighbor of a query in an HNSW Graph. (b): Algorithms [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. 𝑀𝑆𝐸 for a varying number of training queries. 0.95 1.00 Recall 0 5000 # Queries (a) Recall Distr. 0 20000 40000 Dists 0 2000 # Queries (b) Distance Calcs [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: Hyperparameter study, 𝑘 = 50. SIFT100M DEEP100M GLOVE100 GIST1M 0.800.850.900.950.99 Rt 0.80 0.85 0.90 0.95 0.99 Actual Recall (a) Achieved Recalls 0.80 0.85 0.90 0.95 0.99 Rt 1 4 8 12 16 Times Faster (b) Speedup [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 7
Figure 7. Figure 7: Detailed analysis of DARTH for SIFT100M, [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 10
Figure 10. Figure 10: Query recall distribution, 𝑅𝑡 = 0.95, 𝑘 = 50. Baseline LAET REM DARTH 1 10 20 30 Noise % 0.0 0.5 1.0 Actual Recall (a) SIFT100M 1 10 20 30 Noise % 0.0 0.5 1.0 Actual Recall (b) DEEP100M 1 10 20 30 Noise % 0.0 0.5 1.0 Actual Recall (c) GLOVE1M 1 10 20 30 Noise % 0.0 0.…
Figure 11
Figure 11. Figure 11: Recall for varying noise, 𝑅𝑡 = 0.90, 𝑘 = 50. The red line indicates the maximum attainable recall from the plain HNSW index. 4.2.7 Competitor Robustness for Hard Queries. One of the major advantages of DARTH as a run￾time adaptive approach is that it can adapt the ter…
Figure 15
Figure 15. Figure 15: P99. 0.80 0.85 0.90 Rt 0.0 0.5 Worst 1% (a) SIFT100M 0.80 0.85 0.90 Rt 0.0 0.5 Worst 1% (b) DEEP100M 0.80 0.85 0.90 Rt 0.0 0.5 Worst 1% (c) GLOVE1M 0.80 0.85 0.90 Rt 0.0 0.5 Worst 1% (d) GIST1M [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]
Figure 20
Figure 20. Figure 20: Competitor comparison on T2I100M OOD queries (no noise), [PITH_FULL_IMAGE:figures/full_fig_p021_20.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. E2E: Efficient Filtered AKNN Search via Adaptive Termination

    cs.DB 2026-02 conditional novelty 5.0 of 10

    A learned model predicts filtered AKNN search cost from early-probe local filter statistics, enabling per-query early termination with reported speedups of up to ~3x at similar recall.

Reference graph

Works this paper leans on

105 extracted references · 56 canonical work pages · cited by 1 Pith paper

  1. [1]

    pgvector

    2024. pgvector. https://github.com/pgvector/pgvector

    2024

  2. [2]

    DARTH Artifacts

    2025. DARTH Artifacts. https://github.com/MChatzakis/DARTH

    2025

  3. [3]

    Google AI. 2023. Gemini: A Large Language Model. https://geminilang.google

    2023

  4. [4]

    Amazon Web Services. [n. d.]. Amazon Aurora PostgreSQL. https://aws.amazon.com/rds/aurora-postgresql/. Accessed: 2024-11-26

    2024

  5. [5]

    ANN Benchmarks. [n. d.]. ANN Benchmarks HNSW parameters. https://github.com/erikbern/ann-benchmarks/blob/ main/ann_benchmarks/algorithms/hnswlib/config.yml

  6. [6]

    Jason Ansel, Shoaib Kamil, Kalyan Veeramachaneni, Jonathan Ragan-Kelley, Jeffrey Bosboom, Una-May O’Reilly, and Saman Amarasinghe. 2014. Opentuner: An extensible framework for program autotuning. In Proceedings of the 23rd international conference on Parallel architectures and compilation . 303–316. Proc. ACM Manag. Data, Vol. 3, No. 4 (SIGMOD), Article 2...

    2014

  7. [7]

    Martin Aumüller, Erik Bernhardsson, and Alexander Faithfull. 2020. ANN-Benchmarks: A benchmarking tool for approximate nearest neighbor algorithms. Information Systems 87 (2020), 101374

    2020

  8. [8]

    Martin Aumüller and Matteo Ceccarello. 2021. The role of local dimensionality measures in benchmarking nearest neighbor search. Information Systems 101 (2021), 101807

    2021

Show all 105 references
  1. [9]

    Ilias Azizi, Karima Echihabi, and Themis Palpanas. 2023. Elpis: Graph-based similarity search for scalable data science. Proceedings of the VLDB Endowment 16, 6 (2023), 1548–1559

    2023

  2. [10]

    Ilias Azizi, Karima Echihabi, and Themis Palpanas. 2025. Graph-Based Vector Search: An Experimental Evaluation of the State-of-the-Art. PACMMOD (2025)

    2025

  3. [11]

    Artem Babenko and Victor Lempitsky. 2016. Efficient indexing of billion-scale datasets of deep descriptors. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2055–2063

    2016

  4. [12]

    Alexei Botchkarev. 2018. Performance metrics (error measures) in machine learning regression, forecasting and prognostics: Properties and typology. arXiv preprint arXiv:1809.03006 (2018)

    2018 arXiv

  5. [13]

    A Colin Cameron and Frank AG Windmeijer. 1997. An R-squared measure of goodness of fit for some common nonlinear regression models. Journal of econometrics 77, 2 (1997), 329–342

    1997

  6. [14]

    Alessandro Camerra, Themis Palpanas, Jin Shieh, and Eamonn Keogh. 2010. isax 2.0: Indexing and mining one billion time series. In 2010 IEEE international conference on data mining . IEEE, 58–67

    2010

  7. [15]

    Matteo Ceccarello, Alexandra Levchenko, Ileana Ioana, and Themis Palpanas. 2025. Evaluating and Generating Query Workloads for High Dimensional Vector Similarity Search. ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) (2025)

    2025

  8. [16]

    Miriam Cha, Youngjune Gwon, and HT Kung. 2017. Language modeling by clustering with word embeddings for text readability assessment. In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management . 2003–2006

    2017

  9. [17]

    Manos Chatzakis, Panagiota Fatourou, Eleftherios Kosmas, Themis Palpanas, and Botao Peng. 2023. Odyssey: A Journey in the Land of Distributed Data Series Similarity Search. Proc. VLDB Endow. 16, 5 (Jan. 2023), 1140–1153. https://doi.org/10.14778/3579075.3579087

    2023

  10. [18]

    Manos Chatzakis, Michalis Mountantonakis, and Yannis Tzitzikas. 2021. RDFSIM: similarity-based browsing over dbpedia using embeddings. Information 12, 11 (2021), 440

    2021

  11. [19]

    Qi Chen, Haidong Wang, Mingqin Li, Gang Ren, Scarlett Li, Jeffery Zhu, Jason Li, Chuanjie Liu, Lintao Zhang, and Jingdong Wang. 2018. SPTAG: A library for fast approximate nearest neighbor search . https://github.com/Microsoft/ SPTAG

    2018

  12. [20]

    Qi Chen, Bing Zhao, Haidong Wang, Mingqin Li, Chuanjie Liu, Zengzhong Li, Mao Yang, and Jingdong Wang. 2021. Spann: Highly-efficient billion-scale approximate nearest neighborhood search. Advances in Neural Information Processing Systems 34 (2021), 5199–5212

    2021

  13. [21]

    Rihan Chen, Bin Liu, Han Zhu, Yaoxuan Wang, Qi Li, Buting Ma, Qingbo Hua, Jun Jiang, Yunlong Xu, Hongbo Deng, et al. 2022. Approximate nearest neighbor search under neural similarity metric for large-scale recommendation. In Proceedings of the 31st ACM International Conference...

    2022

  14. [22]

    Tianqi Chen and Carlos Guestrin. 2016. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining . 785–794

    2016

  15. [23]

    Google Cloud. 2024. AlloyDB for PostgreSQL. https://cloud.google.com/alloydb/docs/overview Accessed: 2024-12-22

    2024

  16. [24]

    Abhinandan S Das, Mayur Datar, Ashutosh Garg, and Shyam Rajaram. 2007. Google news personalization: scalable online collaborative filtering. In Proceedings of the 16th international conference on World Wide Web . 271–280

    2007

  17. [25]

    Anirban Dasgupta, Ravi Kumar, and Tamás Sarlós. 2011. Fast locality-sensitive hashing. In Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining . 1073–1081

    2011

  18. [26]

    Samuel Daulton, Maximilian Balandat, and Eytan Bakshy. 2020. Differentiable expected hypervolume improvement for parallel multi-objective Bayesian optimization. Advances in Neural Information Processing Systems 33 (2020), 9851–9864

    2020

  19. [27]

    Wei Dong, Charikar Moses, and Kai Li. 2011. Efficient k-nearest neighbor graph construction for generic similarity measures. In Proceedings of the 20th international conference on World wide web . 577–586

    2011

  20. [28]

    Ishita Doshi, Dhritiman Das, Ashish Bhutani, Rajeev Kumar, Rushi Bhatt, and Niranjan Balasubramanian. 2020. LANNS: a web-scale approximate nearest neighbor lookup system. arXiv preprint arXiv:2010.09426 (2020)

    2020 arXiv

  21. [29]

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2024. The faiss library. arXiv preprint arXiv:2401.08281 (2024)

    2024 arXiv

  22. [30]

    Karima Echihabi, Panagiota Fatourou, Kostas Zoumpatianos, Themis Palpanas, and Houda Benbrahim. 2022. Hercules against data series similarity search. arXiv preprint arXiv:2212.13297 (2022)

    2022 arXiv

  23. [31]

    Karima Echihabi, Theophanis Tsandilas, Anna Gogolou, Anastasia Bezerianos, and Themis Palpanas. 2023. ProS: data series progressive k-NN similarity search and classification with probabilistic quality guarantees. The VLDB Journal 32, 4 (2023), 763–789. Proc. ACM Manag. Data, V...

    2023

  24. [32]

    Karima Echihabi, Kostas Zoumpatianos, and Themis Palpanas. 2020. Scalable Machine Learning on High-Dimensional Vectors: From Data Series to Deep Network Embeddings. In International Conference on Web Intelligence, Mining and Semantics WIMS. 1–6

    2020

  25. [33]

    Karima Echihabi, Kostas Zoumpatianos, Themis Palpanas, and Houda Benbrahim. 2020. Return of the lernaean hydra: Experimental evaluation of data series approximate similarity search. arXiv preprint arXiv:2006.11459 (2020)

    2020 arXiv

  26. [34]

    Elastic. [n. d.]. Elasticsearch. https://www.elastic.co/. Accessed: 2024-11-26

    2024

  27. [35]

    Panagiota Fatourou, Eleftherios Kosmas, Themis Palpanas, and George Paterakis. 2023. FreSh: A Lock-Free Data Series Index. In SRDS

    2023

  28. [36]

    Hakan Ferhatosmanoglu, Ertem Tuncel, Divyakant Agrawal, and Amr El Abbadi. 2001. Approximate nearest neighbor searching in multimedia databases. In Proceedings 17th International Conference on Data Engineering . IEEE, 503–511

    2001

  29. [37]

    Jerome H Friedman. 2001. Greedy function approximation: a gradient boosting machine. Annals of statistics (2001), 1189–1232

    2001

  30. [38]

    Jerome H Friedman. 2002. Stochastic gradient boosting. Computational statistics & data analysis 38, 4 (2002), 367–378

    2002

  31. [39]

    Cong Fu, Chao Xiang, Changxu Wang, and Deng Cai. 2017. Fast approximate nearest neighbor search with the navigating spreading-out graph. arXiv preprint arXiv:1707.00143 (2017)

    2017 arXiv

  32. [40]

    Jianyang Gao and Cheng Long. 2024. RaBitQ: Quantizing High-Dimensional Vectors with a Theoretical Error Bound for Approximate Nearest Neighbor Search. Proceedings of the ACM on Management of Data 2, 3 (2024), 1–27

    2024

  33. [41]

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, and Haofen Wang. 2023. Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997 (2023)

    2023 arXiv

  34. [42]

    GASS. [n. d.]. GASS HNSW parameters. https://github.com/zeraph6/GASS_Repo/blob/main/code/README.md

  35. [43]

    Tiezheng Ge, Kaiming He, Qifa Ke, and Jian Sun. 2013. Optimized product quantization. IEEE transactions on pattern analysis and machine intelligence 36, 4 (2013), 744–755

    2013

  36. [44]

    Anna Gogolou, Theophanis Tsandilas, Themis Palpanas, and Anastasia Bezerianos. 2019. Progressive similarity search on time series data. In BigVis 2019-2nd International Workshop on Big Data Visual Exploration and Analytics

    2019

  37. [45]

    Siddharth Gollapudi, Neel Karia, Varun Sivashankar, Ravishankar Krishnaswamy, Nikit Begwani, Swapnil Raz, Yiyong Lin, Yin Zhang, Neelam Mahapatro, Premkumar Srinivasan, et al . 2023. Filtered-diskann: Graph algorithms for approximate nearest neighbor search with filters. In Pr...

    2023

  38. [46]

    Google Cloud. [n. d.]. Vertex AI. https://cloud.google.com/vertex-ai/docs/vector-search/overview. Accessed: 2024-12-19

    2024

  39. [47]

    Yutong Gou, Jianyang Gao, Yuexuan Xu, and Cheng Long. 2024. SymphonyQG: Towards Symphonious Integration of Quantization and Graph for Approximate Nearest Neighbor Search. arXiv preprint arXiv:2411.12229 (2024)

    2024 arXiv

  40. [48]

    Ruiqi Guo, Philip Sun, Erik Lindgren, Quan Geng, David Simcha, Felix Chern, and Sanjiv Kumar. 2020. Accelerating large-scale inference with anisotropic vector quantization. In International Conference on Machine Learning . PMLR, 3887–3896

    2020

  41. [49]

    Yikun Han, Chunjiang Liu, and Pengfei Wang. 2023. A comprehensive survey on vector database: Storage and retrieval technique, challenge. arXiv preprint arXiv:2310.11703 (2023)

    2023

  42. [50]

    Jui-Ting Huang, Ashish Sharma, Shuying Sun, Li Xia, David Zhang, Philip Pronin, Janani Padmanabhan, Giuseppe Ottaviano, and Linjun Yang. 2020. Embedding-based retrieval in facebook search. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & ...

    2020

  43. [51]

    Qiang Huang, Jianlin Feng, Yikai Zhang, Qiong Fang, and Wilfred Ng. 2015. Query-aware locality-sensitive hashing for approximate nearest neighbor search. Proceedings of the VLDB Endowment 9, 1 (2015), 1–12

    2015

  44. [52]

    Shikhar Jaiswal, Ravishankar Krishnaswamy, Ankit Garg, Harsha Vardhan Simhadri, and Sheshansh Agrawal. 2022. Ood-diskann: Efficient and scalable graph anns for out-of-distribution queries. arXiv preprint arXiv:2211.12850 (2022)

    2022 arXiv

  45. [53]

    Daniel Jasbick, Lucio Santos, Paulo M Azevedo-Marques, Agma JM Traina, Daniel de Oliveira, and Marcos Bedo. 2023. Pushing diversity into higher dimensions: The LID effect on diversified similarity searching. Information Systems 114 (2023), 102166

    2023

  46. [54]

    Suhas Jayaram Subramanya, Fnu Devvrit, Harsha Vardhan Simhadri, Ravishankar Krishnawamy, and Rohan Kadekodi

  47. [55]

    Herve Jegou, Matthijs Douze, and Cordelia Schmid. 2010. Product quantization for nearest neighbor search. IEEE transactions on pattern analysis and machine intelligence 33, 1 (2010), 117–128

    2010

  48. [56]

    Hervé Jégou, Romain Tavenard, Matthijs Douze, and Laurent Amsaleg. 2011. Searching in one billion vectors: re-rank with source coding. In 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 861–864

    2011

  49. [57]

    Zhi Jing, Yongye Su, Yikun Han, Bo Yuan, Haiyun Xu, Chunjiang Liu, Kehai Chen, and Min Zhang. 2024. When large language models meet vector databases: a survey. arXiv preprint arXiv:2402.01763 (2024). Proc. ACM Manag. Data, Vol. 3, No. 4 (SIGMOD), Article 242. Publication date:...

    2024 arXiv

  50. [58]

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems 30 (2017)

    2017

  51. [59]

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, et al. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing S...

    2020

  52. [60]

    Conglong Li, Minjia Zhang, David G Andersen, and Yuxiong He. 2020. Improving approximate nearest neighbor search through learned adaptive early termination. In Proceedings of the 2020 ACM SIGMOD International Conference on Management of Data . 2539–2554

    2020

  53. [61]

    Huayang Li, Yixuan Su, Deng Cai, Yan Wang, and Lemao Liu. 2022. A survey on retrieval-augmented text generation. arXiv preprint arXiv:2202.01110 (2022)

    2022 arXiv

  54. [62]

    Yu A Malkov and Dmitry A Yashunin. 2018. Efficient and robust approximate nearest neighbor search using hierarchical navigable small world graphs. IEEE transactions on pattern analysis and machine intelligence 42, 4 (2018), 824–836

    2018

  55. [63]

    Yusuke Matsui, Yusuke Uchida, Hervé Jégou, and Shin’ichi Satoh. 2018. A survey of product quantization. ITE Transactions on Media Technology and Applications 6, 1 (2018), 2–10

    2018

  56. [64]

    Microsoft. [n. d.]. Vectors in Azure AI Search. https://learn.microsoft.com/en-us/azure/search/vector-search-overview. Accessed: 2024-11-26

    2024

  57. [65]

    Microsoft Azure. [n. d.]. Azure Cosmos DB. https://learn.microsoft.com/en-us/azure/cosmos-db/vector-database. Accessed: 2024-12-19

    2024

  58. [66]

    MongoDB, Inc. [n. d.]. MongoDB. https://www.mongodb.com/. Accessed: 2024-12-19

    2024

  59. [67]

    Alexey Natekin and Alois Knoll. 2013. Gradient boosting machines, a tutorial. Frontiers in neurorobotics 7 (2013), 21

    2013

  60. [68]

    OpenAI. 2024. ChatGPT (November 2024 version). https://openai.com Accessed: 2024-11-30

    2024

  61. [69]

    Oracle Corporation. [n. d.]. Oracle AI Vector Search. https://www.oracle.com/database/ai-vector-search/. Accessed: 2024-11-26

    2024

  62. [70]

    Themis Palpanas. 2020. Evolution of a Data Series Index: The iSAX Family of Data Series Indexes: iSAX, iSAX2. 0, iSAX2+, ADS, ADS+, ADS-Full, ParIS, ParIS+, MESSI, DPiSAX, ULISSE, Coconut-Trie/Tree, Coconut-LSM. In Information Search, Integration, and Personalization: 13th Int...

    2020

  63. [71]

    James Jie Pan, Jianguo Wang, and Guoliang Li. 2024. Survey of vector database management systems. The VLDB Journal 33, 5 (2024), 1591–1615

    2024

  64. [72]

    Marco Patella and Paolo Ciaccia. 2008. The many facets of approximate similarity search. In First International Workshop on Similarity Search and Applications (sisap 2008) . IEEE, 10–21

    2008

  65. [73]

    Botao Peng, Panagiota Fatourou, and Themis Palpanas. 2020. Messi: In-memory data series indexing. In 2020 IEEE 36th International Conference on Data Engineering (ICDE) . IEEE, 337–348

    2020

  66. [74]

    Botao Peng, Panagiota Fatourou, and Themis Palpanas. 2021. SING: Sequence indexing using GPUs. In 2021 IEEE 37th International Conference on Data Engineering (ICDE) . IEEE, 1883–1888

    2021

  67. [75]

    Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. Glove: Global vectors for word representation. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP) . 1532–1543

    2014

  68. [76]

    Pinecone, Inc. [n. d.]. Pinecone. https://www.pinecone.io/. Accessed: 2024-12-19

    2024

  69. [77]

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan Hulikal Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Tran, Jonah Samost, et al. 2023. Recommender systems with generative retrieval. Advances in Neural Information Processing Systems 36 (2023), 10299–10315

    2023

  70. [78]

    Jie Ren, Minjia Zhang, and Dong Li. 2020. Hm-ann: Efficient billion-point nearest neighbor search on heterogeneous memory. Advances in Neural Information Processing Systems 33 (2020), 10672–10684

    2020

  71. [79]

    Jason D Rights and Sonya K Sterba. 2019. Quantifying explained variance in multilevel models: An integrative framework for defining R-squared measures. Psychological methods 24, 3 (2019), 309

    2019

  72. [80]

    Viktor Sanca, Manos Chatzakis, and Anastasia Ailamaki. 2024. Optimizing Context-Enhanced Relational Joins. (2024), 501–515. https://doi.org/10.1109/ICDE60146.2024.00045

    2024

  73. [81]

    Min Shi, Jianxun Liu, Dong Zhou, Mingdong Tang, and Buqing Cao. 2017. WE-LDA: a word embeddings augmented LDA model for web services clustering. In 2017 ieee international conference on web services (icws) . IEEE, 9–16

    2017

  74. [82]

    Harsha Vardhan Simhadri, George Williams, Martin Aumüller, Matthijs Douze, Artem Babenko, Dmitry Baranchuk, Qi Chen, Lucas Hosseini, Ravishankar Krishnaswamny, Gopal Srinivasa, et al. 2022. Results of the NeurIPS’21 challenge on billion-scale approximate nearest neighbor searc...

    2022

  75. [83]

    SeMI Technologies. 2019. Weaviate: Open-Source Vector Search Engine. https://weaviate.io. Accessed: 2025-01-15

    2019

  76. [84]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models. Proc. ACM Manag. Data, Vol. 3, No. 4 (SIGMOD), Art...

    2023 arXiv

  77. [85]

    Dana Van Aken, Andrew Pavlo, Geoffrey J Gordon, and Bohan Zhang. 2017. Automatic database management system tuning through large-scale machine learning. In Proceedings of the 2017 ACM international conference on management of data. 1009–1024

    2017

  78. [86]

    Jianguo Wang, Xiaomeng Yi, Rentong Guo, Hai Jin, Peng Xu, Shengjun Li, Xiangyu Wang, Xiangzhou Guo, Chengming Li, Xiaohai Xu, et al . 2021. Milvus: A purpose-built vector data management system. In Proceedings of the 2021 International Conference on Management of Data . 2614–2627

    2021

  79. [87]

    Mengzhao Wang, Xiaoliang Xu, Qiang Yue, and Yuxiang Wang. 2021. A comprehensive survey and experimental comparison of graph-based approximate nearest neighbor search. arXiv preprint arXiv:2101.12631 (2021)

    2021 arXiv

  80. [88]

    Qitong Wang, Ioana Ileana, and Themis Palpanas. 2025. LeaFi: Data Series Indexes on Steroids with Learned Filters. Proc. ACM Manag. Data (2025)

    2025

  81. [89]

    Zeyu Wang, Peng Wang, Themis Palpanas, and Wei Wang. 2023. Graph-and Tree-based Indexes for High-dimensional Vector Similarity Search: Analyses, Comparisons, and Future Directions. IEEE Data Eng. Bull. 46, 3 (2023), 3–21

    2023

  82. [90]

    Zeyu Wang, Qitong Wang, Xiaoxing Cheng, Peng Wang, Themis Palpanas, and Wei Wang. 2024. Steiner-Hardness: A Query Hardness Measure for Graph-Based ANN Indexes. Proceedings of the VLDB Endowment (PVLDB) Journal (2024)

    2024

  83. [91]

    Zeyu Wang, Qitong Wang, Peng Wang, Themis Palpanas, and Wei Wang. 2023. Dumpy: A compact and adaptive index for large data series collections. Proceedings of the ACM on Management of Data 1, 1 (2023), 1–27

    2023

  84. [92]

    Zeyu Wang, Qitong Wang, Peng Wang, Themis Palpanas, and Wei Wang. 2024. DumpyOS: A data-adaptive multi-ary index for scalable data series similarity search. The VLDB Journal 33, 6 (2024), 1887–1911

    2024

  85. [93]

    Zeyu Wang, Haoran Xiong, Qitong Wang, Zhenying He, Peng Wang, Themis Palpanas, and Wei Wang. 2024. Dimensionality-Reduction Techniques for Approximate Nearest Neighbor Search: A Survey and Evaluation. IEEE Data Eng. Bull. 48, 3 (2024), 63–80

    2024

  86. [94]

    WEAVESS. [n. d.]. WEAVESS HNSW parameters. https://github.com/Lsyhprum/WEAVESS/tree/dev/parameters

  87. [95]

    Chuangxian Wei, Bin Wu, Sheng Wang, Renjie Lou, Chaoqun Zhan, Feifei Li, and Yuanzhe Cai. 2020. AnalyticDB-V: a hybrid analytical engine towards query fusion for structured and unstructured data. Proceedings of the VLDB Endowment 13, 12 (2020), 3152–3165

    2020

  88. [96]

    Jiuqi Wei, Xiaodong Lee, Zhenyu Liao, Themis Palpanas, and Botao Peng. 2025. Subspace Collision: An Efficient and Accurate Framework for High-dimensional Approximate Nearest Neighbor Search. PACMMOD (2025)

    2025

  89. [97]

    Jiuqi Wei, Botao Peng, Xiaodong Lee, and Themis Palpanas. 2024. DET-LSH: A Locality-Sensitive Hashing Scheme with Dynamic Encoding Tree for Approximate Nearest Neighbor Search. Proc. VLDB Endow. 17, 9 (2024), 2241–2254

    2024

  90. [98]

    Xingrui Xie, Han Liu, Wenzhe Hou, and Hongbin Huang. 2023. A Brief Survey of Vector Databases. In 2023 9th International Conference on Big Data and Information Analytics (BigDIA) . IEEE, 364–371

    2023

  91. [99]

    Djamel Edine Yagoubi, Reza Akbarinia, Florent Masseglia, and Themis Palpanas. 2017. DPiSAX: Massively Distributed Partitioned iSAX. In IEEE International Conference on Data Mining, ICDM . IEEE Computer Society, 1135–1140

    2017

  92. [100]

    Djamel Edine Yagoubi, Reza Akbarinia, Florent Masseglia, and Themis Palpanas. 2020. Massively Distributed Time Series Indexing and Querying. IEEE Trans. Knowl. Data Eng. 32, 1 (2020), 108–120

    2020

  93. [101]

    Tiannuo Yang, Wen Hu, Wangqi Peng, Yusen Li, Jianguo Li, Gang Wang, and Xiaoguang Liu. 2024. VDTuner: Automated Performance Tuning for Vector Data Management Systems. arXiv preprint arXiv:2404.10413 (2024)

    2024 arXiv

  94. [102]

    Hangjun Ye and Guangyou Xu. 2003. Fast search in large-scale image database using vector quantization. In International Conference on Image and Video Retrieval . Springer, 477–487

    2003

  95. [103]

    Kostas Zoumpatianos, Yin Lou, Ioana Ileana, Themis Palpanas, and Johannes Gehrke. 2018. Generating data series query workloads. The VLDB Journal 27 (2018), 823–846

    2018

  96. [104]

    Kostas Zoumpatianos, Yin Lou, Themis Palpanas, and Johannes Gehrke. 2015. Query workloads for data series indexes. In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . 1603–1612. Received January 2025; revised April 2025; acce...

    2015

  97. [2019]

    Advances in Neural Information Processing Systems 32 (2019)

    Diskann: Fast accurate billion-point nearest neighbor search on a single node. Advances in Neural Information Processing Systems 32 (2019)

    2019

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.