pith. sign in

arxiv: 2607.02046 · v1 · pith:FQKKBZPDnew · submitted 2026-07-02 · 💻 cs.LG

Fast and Accurate Anomaly Detection in Time Series

Pith reviewed 2026-07-03 17:13 UTC · model grok-4.3

classification 💻 cs.LG
keywords anomaly detectiontime seriesHaar wavelett-testunsupervised learningwavelet transformoutlier detectionstatistical testing
0
0 comments X

The pith

A custom t-test on Haar wavelet coefficients detects time series anomalies in an unsupervised way.

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

The paper presents an unsupervised algorithm that applies a specially designed t-test to the coefficients produced by the Haar discrete wavelet transform. It derives the theoretical basis for this test and reports that the resulting detector outperforms existing unsupervised and self-supervised methods when evaluated on 343 datasets. The approach targets the common difficulties of rare anomalies and missing labels by relying only on statistical testing rather than training. A reader would care because the method promises lower false-positive rates in domains where mistakes carry high costs, such as finance or industrial monitoring. If the claim holds, it supplies a fast, label-free alternative whose performance does not depend on post-hoc parameter adjustments.

Core claim

The paper establishes the theoretical foundation of a t-test constructed to operate on Haar discrete wavelet coefficients and shows that the resulting unsupervised detector achieves higher accuracy than current benchmarks across 343 time-series datasets while remaining computationally efficient.

What carries the argument

A custom t-test applied directly to Haar wavelet coefficients that flags anomalies through statistical deviation testing in the transformed domain.

If this is right

  • The detector can be used in safety-critical settings without requiring labelled anomaly examples.
  • Computation remains linear in the series length because the Haar transform and t-test are both fast.
  • No dataset-specific tuning is needed to achieve the reported performance levels.
  • The method scales to large numbers of series because it avoids any training phase.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same coefficient-testing idea could be tried with other wavelet families or with short-time Fourier transforms to handle different frequency characteristics.
  • Online variants might update the t-statistic incrementally as new observations arrive, enabling real-time monitoring.
  • Combining the wavelet test with simple thresholding on raw values could reduce misses on anomalies that are not well captured in the detail coefficients.

Load-bearing premise

The custom t-test retains its statistical validity and power when the input consists of Haar wavelet coefficients from real-world time series that may deviate from ideal theoretical conditions.

What would settle it

A new collection of time-series datasets on which the algorithm fails to match or exceed the accuracy of the strongest published unsupervised and self-supervised baselines would falsify the performance claim.

Figures

Figures reproduced from arXiv: 2607.02046 by Angelo Coluccia, Emanuele Mele, Italo Epicoco, Massimo Cafaro.

Figure 1
Figure 1. Figure 1: Multi-level Haar Wavelet decomposition (𝐿̂ = 3) of a representative signal from the NAB dataset collection. The top panel illustrates the original univariate time series. The subsequent rows display the iterative decomposition process: The left column shows the approximation or coarse coefficients, which capture the low-frequency trend, while the right column isolates the high-frequency detail coefficients… view at source ↗
Figure 2
Figure 2. Figure 2: Example of outlier tree propagation. Starting from the lowest level 𝐿̂ , the detected outliers contribute by setting to one the corresponding leaves (red circles). Level 𝐿̂ − 1 has the same detection behaviour of 𝐿̂ , while retaining the information from the previous level (blue arrows). [𝑎 ∶ 𝑏] follows the convention of being inclusive of the starting index 𝑎 and exclusive of the ending index 𝑏. At each i… view at source ↗
Figure 3
Figure 3. Figure 3: DWTt-test output scores for the “art_load_balancer_spikes” dataset within the NAB collection. • the underlying population follows a normal distribu￾tion, 𝑥𝑖 ∼ (𝜇, 𝜎2 ); • the parameters 𝜇 and 𝜎 2 remain constant throughout the sampling process. In the context of our algorithm, the resulting time se￾ries does not necessarily satisfy these conditions. Tempo￾ral dependencies, auto-correlation, or non-Gaussia… view at source ↗
Figure 4
Figure 4. Figure 4: Statistical analysis of Haar wavelet coefficients for an i.i.d. time series. The raw input signal is constructed such that each observation 𝑥𝑖 is mutually independent and drawn from a target distribution. The plots illustrate the coarse and detail coefficients obtained at a specific decomposition level 𝑙. By applying a moving average with a sliding window of size 𝑤, we extract the local mean behaviour over… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of unsupervised anomaly scores gener￾ated by DWTt-test, DWT-MLEAD, and sLOF. Each subplot illustrates the detection scores overlaid on the original time se￾ries for specific datasets from the SMAP and NAB collections. 4.4.3. AUC-𝑃𝑇 𝑅𝑇 While AUC-PR is robust, its point-wise nature can be excessively punitive for subsequence anomalies. To address this, we also employ range-based Precision (𝑃𝑇 ) an… view at source ↗
Figure 6
Figure 6. Figure 6: Quantitative evaluation of unsupervised algorithms based on AUC-ROC. The line plot (a) shows the metric value achieved for each dataset index, while the boxplot (b) provides a statistical summary of the score distribution across the entire collection. is able to detect the first whole anomaly window while DWT-MLEAD detects only its starting points. Furthermore, the scores computed by the sLOF algorithm hig… view at source ↗
Figure 8
Figure 8. Figure 8: Quantitative evaluation of unsupervised algorithms based on AUC-PTRT. The line plot (a) shows the metric value achieved for each dataset index, while the boxplot (b) provides a statistical summary of the score distribution across the entire collection. as DWTt-test maintains a high throughput while ensuring robust detection capabilities. 4.6. Evaluation of Self-Supervised Algorithms After training the self… view at source ↗
Figure 10
Figure 10. Figure 10: Quantitative evaluation of unsupervised and su￾pervised algorithms based on AUC-ROC. The line plot (a) shows the metric value achieved for each dataset index within the NASA SMAP and MSL collections, while the boxplot (b) provides a statistical summary of the score distribution for these specific sequences. The quantitative summary results are available in Figures 10, 11 and 12. In particular, the latter … view at source ↗
Figure 11
Figure 11. Figure 11: Quantitative evaluation of unsupervised and super￾vised algorithms based on AUC-PR. The line plot (a) shows the metric value achieved for each dataset index within the NASA SMAP and MSL collections, while the boxplot (b) provides a statistical summary of the score distribution for these specific sequences. overall efficiency. However, even with the advantage of high￾end GPU acceleration, our unsupervised … view at source ↗
read the original abstract

Anomaly detection is a critical and evolving field in Machine Learning, with applications targeting different domains such as cybersecurity, finance, healthcare, manufacturing and IoT (Internet of Things) systems. Traditionally, anomaly detection algorithms have been designed using both supervised and unsupervised learning paradigms. The fundamental challenge in real-world anomaly detection scenarios is related to the inherent class imbalance (anomalies are typically rare) and, for supervised methods, to the scarcity of labelled anomalous data. Indeed, labelling is both expensive and time-consuming. Conversely unsupervised methods do not require labelling, but may suffer from high false positive rates when deployed in safety-critical applications. In this work we introduce a novel unsupervised algorithm for anomaly detection in time series based on the Haar discrete wavelet and a suitably designed $t$-test. We establish the theoretical foundation of the proposed $t$-test and, through extensive experimentation across 343 datasets, demonstrate that our algorithm outperforms state-of-the-art unsupervised and self-supervised benchmarks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 1 minor

Summary. The manuscript introduces a novel unsupervised anomaly detection algorithm for time series based on the Haar discrete wavelet transform paired with a custom-designed t-test. The authors claim to establish the theoretical foundation of this t-test and report that extensive experiments across 343 datasets show the method outperforming state-of-the-art unsupervised and self-supervised benchmarks.

Significance. If the theoretical foundation for the t-test holds and the reported performance gains are robust, the work would be significant for providing a fast, label-free approach that mitigates high false-positive rates in safety-critical domains. The scale of the empirical evaluation (343 datasets) is a clear strength that supports generalizability claims.

minor comments (1)
  1. The abstract refers to a 'suitably designed t-test' without previewing key design choices; adding a brief high-level description of the test statistic or its wavelet-specific properties in the abstract or introduction would improve accessibility.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their review. The summary accurately reflects the manuscript's contributions, and we appreciate the recognition of the empirical evaluation scale as a strength. No specific major comments are listed in the report.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The provided abstract and description establish a novel unsupervised method using Haar wavelets and a custom t-test, with a claimed theoretical foundation and empirical validation on 343 datasets. No load-bearing steps reduce by construction to fitted inputs, self-definitions, or self-citation chains; the derivation chain for the t-test and performance claims remains independent of its own outputs. This is the expected honest non-finding for a paper whose central claims rest on external benchmarks rather than internal redefinitions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities. The method appears to rely on standard Haar wavelet properties and a custom t-test whose design details are not supplied.

pith-pipeline@v0.9.1-grok · 5697 in / 1122 out tokens · 24792 ms · 2026-07-03T17:13:04.016959+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 36 canonical work pages

  1. [1]

    Neural Networks and Deep Learning: A Textbook

    Aggarwal, C.C., 2023. Neural Networks and Deep Learning: A Textbook. 2nd ed., Springer Publishing Company, Incorporated. doi:10.1007/978-3-031-29642-0

  2. [2]

    Outlier detection for high dimen- sionaldata,in:Proceedingsofthe2001ACMSIGMODInternational Conference on Management of Data, Association for Computing Machinery, New York, NY, USA

    Aggarwal, C.C., Yu, P.S., 2001. Outlier detection for high dimen- sionaldata,in:Proceedingsofthe2001ACMSIGMODInternational Conference on Management of Data, Association for Computing Machinery, New York, NY, USA. p. 37–46. URL:https://doi.org/ 10.1145/375663.375668, doi:10.1145/375663.375668

  3. [3]

    Unsupervised real- time anomaly detection for streaming data

    Ahmad, S., Lavin, A., Purdy, S., Agha, Z., 2017. Unsupervised real- time anomaly detection for streaming data. Neurocomputing 262. doi:10.1016/j.neucom.2017.04.070

  4. [4]

    Anomaly detection in time series data using reversible instance normalized anomaly transformer

    Baidya, R., Jeong, H., 2023. Anomaly detection in time series data using reversible instance normalized anomaly transformer. Sensors

  5. [5]

    URL:https://www.mdpi.com/1424-8220/23/22/9272, doi:10.3390/ s23229272

  6. [6]

    Automatic outlier detection for timeseries:Anapplicationtosensordata

    Basu, S., Meckesheimer, M., 2007. Automatic outlier detection for timeseries:Anapplicationtosensordata. Knowl.Inf.Syst.11,137–

  7. [7]

    doi:10.1007/s10115-006-0026-6

  8. [8]

    Benali,A.A.E.,Cafaro,M.,Epicoco,I.,Pulimeno,M.,Schioppa,E.J.,

  9. [9]

    IEEE Access 12, 178751–178767

    Just in time transformers. IEEE Access 12, 178751–178767. doi:10.1109/ACCESS.2024.3504862

  10. [10]

    Scaling clustering algo- rithms to large databases, in: Proceedings of the Fourth International ConferenceonKnowledge DiscoveryandDataMining, AAAIPress

    Bradley, P.S., Fayyad, U., Reina, C., 1998. Scaling clustering algo- rithms to large databases, in: Proceedings of the Fourth International ConferenceonKnowledge DiscoveryandDataMining, AAAIPress. p. 9–15

  11. [11]

    Breunig, M.M., Kriegel, H.P., Ng, R.T., Sander, J., 2000a. Lof: identifying density-based local outliers, in: Proceedings of the 2000 ACM SIGMOD International Conference on Management of Data, Association for Computing Machinery, New York, NY, USA. p. 93–104. URL:https://doi.org/10.1145/342009.335388, doi:10.1145/ 342009.335388

  12. [12]

    Lof:iden- tifyingdensity-basedlocaloutliers

    Breunig,M.M.,Kriegel,H.P.,Ng,R.T.,Sander,J.,2000b. Lof:iden- tifyingdensity-basedlocaloutliers. SIGMODRec.29,93–104. URL: https://doi.org/10.1145/335191.335388, doi:10.1145/335191.335388

  13. [13]

    Smote:syntheticminorityover-samplingtechnique

    Chawla, N.V., Bowyer, K.W., Hall, L.O., Kegelmeyer, W.P., 2002. Smote:syntheticminorityover-samplingtechnique. J.Artif.Int.Res. 16, 321–357

  14. [14]

    A k-nearest neigh- bors approach to the design of radar detectors

    Coluccia, A., Fascista, A., Ricci, G., 2020. A k-nearest neigh- bors approach to the design of radar detectors. Signal Processing 174,107609. URL:https://www.sciencedirect.com/science/article/ pii/S0165168420301523, doi:https://doi.org/10.1016/j.sigpro.2020. 107609

  15. [15]

    Hawkins, d

    Enderlein, G., 1987. Hawkins, d. m.: Identification of outliers. chap- man and hall, london – new york 1980, 188 s., £ 14, 50. Biometrical Journal 29, 198–198. doi:https://doi.org/10.1002/bimj.4710290215

  16. [16]

    Ester, M., Kriegel, H.P., Sander, J., Xu, X., 1996. A density-based algorithm for discovering clusters in large spatial databases with noise, in: Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, AAAI Press. p. 226–231

  17. [17]

    Bayesian signal processing for robust iot: Classical vs

    Fascista, A., Coluccia, A., 2025. Bayesian signal processing for robust iot: Classical vs. graph based methods for joint estimation and Emanuele Mele et al.:Preprint submitted to ElsevierPage 17 of 19 Fast and Accurate Anomaly Detection in Time Series anomaly detection. IEEE Internet of Things Magazine 8, 112–119. doi:10.1109/MIOT.2025.3575885

  18. [18]

    Procedures for detecting outlying observations in samples

    Grubbs, F.E., 1969. Procedures for detecting outlying observations in samples. Technometrics 11, 1–21. doi:10.1080/00401706.1969. 10490657

  19. [19]

    Guha, S., Rastogi, R., Shim, K., 1998. Cure: an efficient clustering algorithmforlargedatabases,in:Proceedingsofthe1998ACMSIG- MODInternationalConferenceonManagementofData,Association for Computing Machinery, New York, NY, USA. p. 73–84. URL: https://doi.org/10.1145/276304.276312, doi:10.1145/276304.276312

  20. [20]

    Extendedisolationforest

    Hariri,S.,Kind,M.C.,Brunner,R.J.,2021. Extendedisolationforest. IEEE Transactions on Knowledge and Data Engineering 33, 1479–

  21. [21]

    doi:10.1109/TKDE.2019.2947676

  22. [22]

    He, H., Bai, Y., Garcia, E.A., Li, S., 2008. Adasyn: Adaptive synthetic sampling approach for imbalanced learning, in: 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), pp. 1322–1328. doi:10. 1109/IJCNN.2008.4633969

  23. [23]

    Dasvdd: Deep autoencoding sup- portvectordatadescriptorforanomalydetection

    Hojjati, H., Armanfard, N., 2024. Dasvdd: Deep autoencoding sup- portvectordatadescriptorforanomalydetection. IEEETransactions on Knowledge and Data Engineering 36, 3739–3750. doi:10.1109/ TKDE.2023.3328882

  24. [24]

    Hundman, K., Constantinou, V., Laporte, C., Colwell, I., Soder- strom, T., 2018. Detecting spacecraft anomalies using lstms and nonparametric dynamic thresholding, in: Proceedings of the 24th ACMSIGKDDInternationalConferenceonKnowledgeDiscovery& DataMining,AssociationforComputingMachinery,NewYork,NY, USA. p. 387–395. URL:https://doi.org/10.1145/3219819.32...

  25. [25]

    Ihler,A.,Hutchins,J.,Smyth,P.,2006. Adaptiveeventdetectionwith time-varying poisson processes, in: Proceedings of the 12th ACM SIGKDD International Conference on Knowledge Discovery and DataMining,AssociationforComputingMachinery,NewYork,NY, USA. p. 207–216. URL:https://doi.org/10.1145/1150402.1150428, doi:10.1145/1150402.1150428

  26. [26]

    A review of anomaly detection usingmachinelearningtechniques,in:20241stInternationalConfer- ence on Advanced Computing and Emerging Technologies (ACET), pp

    Kesharwani, A., Shukla, P., 2024. A review of anomaly detection usingmachinelearningtechniques,in:20241stInternationalConfer- ence on Advanced Computing and Emerging Technologies (ACET), pp. 1–6. doi:10.1109/ACET61898.2024.10730574

  27. [27]

    Mamba-3: Improved sequence modeling using state space principles.arXiv preprint arXiv:2603.15569, 2026

    Lahoti, A., Li, K.Y., Chen, B., Wang, C., Bick, A., Kolter, J.Z., Dao, T., Gu, A., 2026. Mamba-3: Improved sequence modeling usingstatespaceprinciples.URL:https://arxiv.org/abs/2603.15569, arXiv:2603.15569

  28. [28]

    Li, D., Chen, D., Jin, B., Shi, L., Goh, J., Ng, S.K., 2019. Mad-gan: Multivariate anomaly detection for time series data with generative adversarial networks, in: Tetko, I.V., Kůrková, V., Karpov, P., Theis, F.(Eds.),ArtificialNeuralNetworksandMachineLearning–ICANN 2019:TextandTimeSeries,SpringerInternationalPublishing,Cham. pp. 703–716

  29. [29]

    A survey of deep learning for industrial visual anomaly detection

    Li, Z., Yan, Y., Wang, X., Ge, Y., Meng, L., 2025. A survey of deep learning for industrial visual anomaly detection. Artif. Intell. Rev. 58, 279. URL:https://doi.org/10.1007/s10462-025-11287-7, doi:10.1007/S10462-025-11287-7

  30. [30]

    A review of deep learning-based anomaly detection strategies in industry 4.0 focused on application fields, sensing equipment, and algorithms

    Liso, A., Cardellicchio, A., Patruno, C., Nitti, M., Ardino, P., Stella, E., Renò, V., 2024. A review of deep learning-based anomaly detection strategies in industry 4.0 focused on application fields, sensing equipment, and algorithms. IEEE Access 12, 93911–93923. doi:10.1109/ACCESS.2024.3424488

  31. [32]

    Isolation forest, in: Proceedings of the 2008 Eighth IEEE International Conference on Data Mining, IEEE Computer Society, USA

    Liu, F.T., Ting, K.M., Zhou, Z.H., 2008b. Isolation forest, in: Proceedings of the 2008 Eighth IEEE International Conference on Data Mining, IEEE Computer Society, USA. p. 413–422. URL: https://doi.org/10.1109/ICDM.2008.17, doi:10.1109/ICDM.2008.17

  32. [33]

    Recurrent neural networks for anomaly detection in magnet power supplies of particle accelerators

    Lobach, I., Borland, M., 2024. Recurrent neural networks for anomaly detection in magnet power supplies of particle accelerators. Machine Learning with Applications 18, 100585. URL:https: //www.sciencedirect.com/science/article/pii/S2666827024000616, doi:https://doi.org/10.1016/j.mlwa.2024.100585

  33. [34]

    Wavelets and operators

    Meyer, Y., 1992. Wavelets and operators. 37, Cambridge university press

  34. [35]

    Deepone-classclassification, in: Dy, J., Krause, A

    Ruff, L., Vandermeulen, R., Goernitz, N., Deecke, L., Siddiqui, S.A., Binder,A.,Müller,E.,Kloft,M.,2018. Deepone-classclassification, in: Dy, J., Krause, A. (Eds.), Proceedings of the 35th International Conference on Machine Learning, PMLR. pp. 4393–4402. URL: https://proceedings.mlr.press/v80/ruff18a.html

  35. [36]

    f-anogan: Fast unsupervised anomaly detection with generative adversarial networks

    Schlegl,T.,Seeböck,P.,Waldstein,S.M.,Langs,G.,Schmidt-Erfurth, U., 2019. f-anogan: Fast unsupervised anomaly detection with generative adversarial networks. Medical Image Analysis 54, 30–44. doi:https://doi.org/10.1016/j.media.2019.01.010

  36. [37]

    Anomaly detection in time series: a comprehensive evaluation

    Schmidl, S., Wenig, P., Papenbrock, T., 2022. Anomaly detection in time series: a comprehensive evaluation. Proc. VLDB Endow. 15, 1779–1797. URL:https://doi.org/10.14778/3538598.3538602, doi:10.14778/3538598.3538602

  37. [38]

    Estimating the support of a high-dimensional dis- tribution

    Schölkopf, B., Platt, J.C., Shawe-Taylor, J., Smola, A.J., Williamson, R.C., 2001. Estimating the support of a high-dimensional dis- tribution. Neural Computation 13, 1443–1471. doi:10.1162/ 089976601750264965

  38. [39]

    Mamba adaptive anomaly transformer with association discrepancy for time series

    Sellam, A.Z., Benaissa, I., Taleb-Ahmed, A., Patrono, L., Distante, C., 2025. Mamba adaptive anomaly transformer with association discrepancy for time series. Engineering Applications of Artifi- cial Intelligence 160, 111685. URL:https://www.sciencedirect. com/science/article/pii/S0952197625016872,doi:https://doi.org/10. 1016/j.engappai.2025.111685

  39. [40]

    A review of deep learning based anomaly detection

    Shi, Y., Yang, G., Wang, X., Yu, Q., Feng, Y., Liu, A., 2026. A review of deep learning based anomaly detection. Neurocomputing 668,132383. URL:https://www.sciencedirect.com/science/article/ pii/S0925231225030553, doi:https://doi.org/10.1016/j.neucom.2025. 132383

  40. [41]

    Sparsetransformer for anomaly detection with association discrepancy, in: 2025 IEEE 37th International Conference on Tools with Artificial Intelligence (ICTAI), pp

    Shrestha,S.,Zhang,Y.,Liu,H.,Silwal,A.,2025. Sparsetransformer for anomaly detection with association discrepancy, in: 2025 IEEE 37th International Conference on Tools with Artificial Intelligence (ICTAI), pp. 591–598. doi:10.1109/ICTAI66417.2025.00086

  41. [42]

    Somepalli, G., Wu, Y., Balaji, Y., Vinzamuri, B., Feizi, S., 2021. Unsupervised anomaly detection with adversarial mirrored autoen- coders,in:deCampos,C.,Maathuis,M.H.(Eds.),Proceedingsofthe Thirty-Seventh Conference on Uncertainty in Artificial Intelligence, PMLR. pp. 365–375. URL:https://proceedings.mlr.press/v161/ somepalli21a.html

  42. [43]

    Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., Pei, D., 2019. Robust anomaly detection for multivariate time series through stochastic recurrentneuralnetwork,in:Proceedingsofthe25thACMSIGKDD International Conference on Knowledge Discovery & Data Mining, Association for Computing Machinery, New York, NY, USA. p. 2828–2837. URL:https://doi.org/10.1145/3292500...

  43. [44]

    Precision and recall for time series, in: Bengio, S., Wallach, H., Larochelle, H., Grauman, K., Cesa-Bianchi, N., Garnett, R

    Tatbul, N., Lee, T.J., Zdonik, S., Alam, M., Gottschlich, J., 2018. Precision and recall for time series, in: Bengio, S., Wallach, H., Larochelle, H., Grauman, K., Cesa-Bianchi, N., Garnett, R. (Eds.), Advances in Neural Information Processing Systems, Curran As- sociates, Inc. URL:https://proceedings.neurips.cc/paper_files/ paper/2018/file/8f468c873a32bb...

  44. [45]

    5281/zenodo.3762385

    Thill,M.,Konen,W.,Bäck,T.,2020.Markusthill/mgab:Themackey- glass anomaly benchmark (version v1.0.1).https://doi.org/10. 5281/zenodo.3762385. URL:https://doi.org/10.5281/zenodo.3762385, doi:10.5281/ZENODO.3762385. accessed on YYYY-MM-DD

  45. [46]

    Time series anomaly detection withdiscretewavelettransformsandmaximumlikelihoodestimation

    Thill, M., Konen, W., Bäck, T., 2019. Time series anomaly detection withdiscretewavelettransformsandmaximumlikelihoodestimation

  46. [47]

    Anomaly detection usinglongshorttermmemorynetworksanditsapplicationsinsupply chain management

    Tran, K.P., Nguyen, H.D., Thomassey, S., 2019. Anomaly detection usinglongshorttermmemorynetworksanditsapplicationsinsupply chain management. IFAC-PapersOnLine 52, 2408–2412. doi:https: //doi.org/10.1016/j.ifacol.2019.11.567. 9th IFAC Conference on Manufacturing Modelling, Management and Control MIM 2019. Emanuele Mele et al.:Preprint submitted to Elsevie...

  47. [48]

    Attention is all you need, in: Proceedings of the 31st International Conference on Neural Infor- mation Processing Systems, Curran Associates Inc., Red Hook, NY, USA

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez, A.N., Kaiser, L., Polosukhin, I., 2017. Attention is all you need, in: Proceedings of the 31st International Conference on Neural Infor- mation Processing Systems, Curran Associates Inc., Red Hook, NY, USA. p. 6000–6010

  48. [49]

    Contrastive deep support vector data description

    Xing, H.J., Zhang, P.P., 2023. Contrastive deep support vector data description. Pattern Recognition 143, 109820. URL:https: //www.sciencedirect.com/science/article/pii/S0031320323005186, doi:https://doi.org/10.1016/j.patcog.2023.109820

  49. [50]

    arXiv preprint arXiv:2110.02642 , year=

    Xu, J., Wu, H., Wang, J., Long, M., 2022a. Anomaly transformer: Time series anomaly detection with association discrepancy. URL: https://arxiv.org/abs/2110.02642,arXiv:2110.02642

  50. [51]

    Anomaly transformer: Time series anomaly detection with association discrepancy, in: In- ternational Conference on Learning Representations

    Xu, J., Wu, H., Wang, J., Long, M., 2022b. Anomaly transformer: Time series anomaly detection with association discrepancy, in: In- ternational Conference on Learning Representations. URL:https: //openreview.net/forum?id=LzQQ89U1qm_

  51. [52]

    Yang, Y., Zhang, C., Zhou, T., Wen, Q., Sun, L., 2023. Dcde- tector: Dual attention contrastive representation learning for time series anomaly detection, in: Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Association for Computing Machinery, New York, NY, USA. p. 3033–3045. URL:https://doi.org/10.1145/3580305.359929...

  52. [53]

    Matrix profile i: All pairs similarity joins for time series: A unifying view that includes motifs, discords and shapelets, pp

    Yeh, C.C.M., Zhu, Y., Ulanova, L., Begum, N., Ding, Y., Dau, A., Silva, D., Mueen, A., Keogh, E., 2016. Matrix profile i: All pairs similarity joins for time series: A unifying view that includes motifs, discords and shapelets, pp. 1317–1322. doi:10.1109/ICDM.2016.0179

  53. [54]

    Time series outlier detection based on sliding window prediction

    yf, y., Zhu, Y., Li, S., Wan, D., 2014. Time series outlier detection based on sliding window prediction. Mathematical Problems in Engineering 2014. doi:10.1155/2014/879736. Emanuele Mele et al.:Preprint submitted to ElsevierPage 19 of 19