Decomposition-based multi-scale transformer framework for time series anomaly detection

Cuicui Luo; Wenxin Zhang

REVIEW 5 major objections 5 minor 13 references

Decomposition-based multi-scale transformer framework for time series anomaly detection

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read TransDe claims that decomposing a time series into trend and cycle, then aligning intra-patch and inter-patch transformer views with KL divergence, yields leading F1 scores on four multivariate anomaly benchmarks.

desk verdict Plausible incremental architecture with competitive F1 results on four of five benchmarks, but the written method is internally inconsistent and needs substantial revision before it can be evaluated. read the letter →

arxiv 2504.14206 v1 pith:5KOUJEZ4 submitted 2025-04-19 cs.LG cs.AI

Wenxin Zhang , Cuicui Luo This is my paper · ORCID

classification cs.LGcs.AI

keywords timeseriesanomalydetectionunsupervisedlearningtransformerdecompositioncontrastiveKLdivergencepatch-basedrepresentationmultivariate

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

TransDe, the framework proposed here, is an unsupervised multivariate time-series anomaly detector that combines three moves: Hodrick-Prescott decomposition into trend and cyclical components, multi-scale patching with a shared-weight transformer that computes intra-patch and inter-patch attention views, and a contrastive loss that aligns these views with symmetric KL divergence under a stop-gradient. The paper claims this combination learns normal patterns well enough that anomalies, which produce inconsistent views, receive high anomaly scores. On four public benchmarks (MSL, SMAP, SWaT, and PSM), the reported F1 scores are 94.36, 96.67, 98.04, and 96.43, ahead of twelve baselines; on SMD the method trails the best baseline, which the authors attribute to single-point anomalies and to normal points being misread as anomalous. The central assertion is that decomposition plus contrastive patch views, rather than reconstruction or prediction error, is the right objective for anomaly scoring under noise.

What carries the argument

The load-bearing object is the pair of patch-level attention views, $\hat{\Omega}_{\mathrm{intra}}$ and $\hat{\Omega}_{\mathrm{inter}}$, produced from HP-filtered trend and cyclical components and expanded to the same $\mathbb{R}^{T\times T}$ dimension. The Hodrick-Prescott filter is a smoothing procedure that splits a series $x_t$ into a trend $\tau_t$ and a cyclical residual $\varepsilon_t$; this is the decomposition that feeds the multi-scale patch channels. The paper uses the symmetric KL divergence between the two views as a stand-in for pattern consistency: normal patterns are supposed to agree across views, so divergence stays low, while rare anomalous patterns are supposed to disagree, so divergence rises. A stop-gradient operation on the target view makes the loss asymmetric in update flow, which the paper claims prevents model collapse and improves the learned embeddings, and multi-scale patch sizes act as independent channels whose losses are combined.

What would settle it

Train TransDe on normal data only, then feed a held-out normal segment containing legitimate abrupt changes such as sensor recalibration, mode switches, or level shifts, and count how often the KL score crosses the threshold $\rho$; if many normal points score above the threshold, the underlying consistency assumption is false.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is a detection pipeline in which the anomaly score is the KL divergence between two transformer-generated views of the same decomposed, patched sequence. The trend and cyclical components are segmented at several patch sizes; one attention pass looks inside patches (intra-patch dependencies, a $\mathbb{R}^{P\times P}$ matrix) and another looks across patches (inter-patch dependencies, a $\mathbb{R}^{N\times N}$ matrix). These representations are expanded to a common $\mathbb{R}^{T\times T}$ size, fused across components, and scored by the symmetric KL divergence between the two views, with a stop-gradient applied to the target side; the same loss is used for training and for the test-time anomaly score. The reported experiments give best F1 values of 94.36 on MSL, 96.67 on SMAP, 98.04 on SWaT, and 96.43 on PSM, with SMD at 87.33, and the authors attribute the gains to isolating trend and cyclical structure and to avoiding MSE-style optimization that is vulnerable to noise.

Load-bearing premise

The entire detection pipeline rests on the unproven assumption that every anomaly inflates the KL divergence between the two patch views while every normal segment keeps it low; if a normal but unusual regime or a subtle point anomaly violates this, the framework will either raise false alarms or miss anomalies.

Editorial extensions

If this is right

The reported F1 results imply that reconstruction and prediction objectives are not required for strong anomaly detection; a divergence-based score computed directly from two attention views is enough to outperform twelve baselines on four datasets.
Because the loss is computed from positive pairs only, the framework avoids negative-sample mining, and the paper's timing comparison indicates lower training cost per epoch than SimCLR- and MoCo-style contrastive learning.
The ablations show that both inter-patch and intra-patch channels contribute: dropping either reduces F1, so the multi-scale two-view design, not the transformer alone, is doing the work.
The stop-gradient and symmetric-KL choices are load-bearing: ablations in which either is removed degrade performance across datasets.
The framework is not uniformly better: on SMD it falls below the best baseline, and the authors link this to single-point anomalies and to normal signals being misclassified.

Reading between the lines

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

Editorial inference: the paper treats HP decomposition as fixed preprocessing, but the contrastive objective does not depend on the specific filter; swapping in other decomposition schemes would test whether decomposition itself or the HP choice carries the gain.
Editorial inference: the anomaly score is exactly the training loss with stop-gradient, so threshold selection may determine much of the reported F1, and the paper proposes no principled thresholding method.
Editorial inference: since one attention head and shallow encoders perform best, the transformer may be acting as a lightweight alignment module rather than a deep sequence model; replacing it with a linear or convolutional patch projector would be a natural testable extension.
Editorial inference: the SMD weakness and the false positives on normal segments shown in the visualization suggest the core assumption that rare patterns always inflate KL divergence may fail on point anomalies; per-channel thresholds or a prior on expected divergence are natural extensions.

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.

Desk Editor's Note

read the letter

You should know two things up front. First, the paper builds a genuinely new combination — HP-filter decomposition, multi-scale patch transformers, and a stop-gradient symmetric-KL contrastive loss — and on four of five benchmarks it reports F1 gains over several strong baselines. Second, the write-up is not trustworthy as it stands: the equations don't type-check, the implementation contradicts the hyperparameter analysis, and the evaluation reporting is incomplete.

The strongest part is the experiment suite. Five standard datasets, twelve baselines, and ablations on stop-gradient, patch level, loss function, normalization, contrastive paradigm, and hyperparameters. The SMD weakness is acknowledged honestly, which is more than many papers do. If the code matches the intended design, this is a competitive incremental method.

Now the soft spots, in order of severity. Equations (2)–(6) have dimension mismatches: a weight matrix in R^{(d_model/h)×(d_model/h)} multiplied by an embedding in R^{P×d_model} cannot produce Q in R^{N×d_model/h}. These need to be rewritten before anyone can verify the architecture. The stress-test claim that Eq. (12) is identically zero does not hold: L_intra and L_inter are symmetric KL divergences between different argument pairs, so their difference is not zero in general. But the loss is still odd — it subtracts two positive terms and is never clearly motivated — and the anomaly score in Eq. (13) is not the same as the training loss, which is unexplained. There are also concrete internal contradictions: Section 5.4 says two encoder layers while Section 5.8 says the optimum is three; Table 3's patch sizes for SMAP and SWaT differ from the "best" values in Table 8. No error bars are given, threshold selection is not described, and the per-dataset patch/window search looks like tuning on test data. The omission of DCdetector and TFAD is glaring since both are direct predecessors of pieces of this method.

Who is this for? A reader working on contrastive or decomposition-based time series anomaly detection who wants to see one more engineering point on the F1 frontier. It is not a conceptual breakthrough. It deserves a serious referee because the empirical claim is sharp and the architecture is coherent enough to be salvageable, but the current text cannot be accepted as the description of the method that produced the numbers.

Recommendation: send it to review, and tell the authors to fix the equations, reconcile the hyperparameters, and supply error bars and a threshold-selection protocol. If the code checks out, major revision could make it publishable.

Referee Report

5 major / 5 minor

Summary. The paper proposes TransDe, an unsupervised multivariate time series anomaly detection framework that combines Hodrick-Prescott decomposition, multi-scale patch-based transformer encoders, and a KL-divergence contrastive objective with stop-gradient. The authors claim that decomposing the series into trend and cyclical components and learning inter-patch and intra-patch dependencies through contrastive learning yields state-of-the-art F1 scores on five public benchmarks (SMD, MSL, SMAP, SWaT, PSM) compared with twelve baselines. The method is described in Section 4, with experiments in Section 5 that report best F1 values of 87.33, 94.36, 96.67, 98.04, and 96.43 on the five datasets respectively.

Significance. If the proposed framework worked as described, it would be a competitive contribution to time series anomaly detection, particularly in its use of decomposition and contrastive learning without negative samples. The paper includes ablations for stop-gradient, patch-level choices, and loss variants, and it provides a public code link. However, the central methodological derivation contains a degenerate training objective and multiple dimension inconsistencies that make the described model incoherent. The empirical results, while strong on some datasets, are not supported by the written method, and the absence of threshold-selection details and error bars further weakens the evaluation. The core idea is plausible, but the current manuscript does not provide a reproducible or internally consistent method.

major comments (5)

[4.5, Eq. (12)] As written, L_intra and L_inter are the same sum of two KL terms up to commutation: L_intra = KL(Ω̂_intra||Stop(Ω̂_inter)) + KL(Stop(Ω̂_inter)||Ω̂_intra) and L_inter = KL(Ω̂_inter||Stop(Ω̂_intra)) + KL(Stop(Ω̂_intra)||Ω̂_inter) evaluate identically in the forward pass because Stop(·) is the identity. Consequently L = (L_intra − L_inter)/C is identically zero for every input, so Eq. (12) does not define a learning signal. The stop-gradient only changes the backward pass; the paper neither acknowledges this nor explains why optimizing a zero-valued objective with a pseudo-gradient is valid. This is a load-bearing flaw because the training procedure described cannot produce the reported F1 values.
[4.3, Eqs. (2)–(7)] Several dimension mismatches make the attention equations incoherent. In Eq. (2), W_Q, W_K ∈ R^{(d_model/h)×(d_model/h)} are multiplied by X_intra ∈ R^{P×d_model}, but the product is claimed to be Q_intra, K_intra ∈ R^{N×d_model/h}; the inner dimensions d_model/h and d_model do not match, and the output length P is replaced by N. Similarly, Eq. (5) with X_inter ∈ R^{N×d_model} and the same W_Q cannot yield Q_inter ∈ R^{N×d_model/h}. Moreover, Eq. (3) would give Z_intra ∈ R^{N×N} (or P×P depending on the correct shapes), but Eq. (4) asserts Ω_intra ∈ R^{P×P}, with no explanation of how the head dimension and length are reconciled. These inconsistencies mean the encoder is not specified as a computational graph.
[5.5, Table 4 and Abstract] The abstract claims superiority over twelve baselines in terms of F1 score, but Table 4 shows that on SMD, TransDe (F1 = 87.33) is outperformed by NormFAAE (93.82), TranAD (91.57), DCFF-MTAD (91.40), and AnomalyTrans (90.33). The text acknowledges this only later in Section 5.5, which conflicts with the unqualified claim in the abstract and in the contributions list.
[4.6 and 5.4] The anomaly threshold ρ in Eq. (14) is a free hyperparameter, yet the paper does not describe any procedure for setting it (e.g., on a validation set, by quantile, or by grid search). If ρ is selected using test-set labels, the reported F1 values are optimistically biased and cannot be compared with baselines that may have used different threshold strategies. The manuscript should state the threshold selection rule and, ideally, report the dependence of F1 on ρ.
[4.6] The central assumption that anomalies cause a high KL divergence between Ω̂_intra and Ω̂_inter is asserted without derivation or citation: 'any anomalies with rare patterns will result in inconsistent representations, leading to a high anomaly score.' No statistical or empirical argument is given for why normal data must yield small divergence, and because the training loss in Eq. (12) is degenerate, the paper provides no learned-encoder justification for this premise.

minor comments (5)

[5.4 and 5.8] Section 5.4 says TransDe includes two encoder layers, but Section 5.8 concludes the optimal number of layers is 3; this inconsistency affects reproducibility and should be resolved.
[Table 4] Table 4 contains typographical errors: '0.86.37' for OmniAnomaly on MSL and '0.92.45' for IForest on PSM; the SMD column for IForest shows '76.09 0.92.45 83.48' which should likely be '76.09 92.45 83.48'.
[References] The references contain duplicates: Su et al. 2019a and 2019b are the same paper, and Xu et al. 2022a and 2022b are the same paper.
[Section 5.5] No standard deviations or multiple seeds are reported, which makes the claimed improvements of 0.11% and 0.26% on MSL and SMAP hard to assess.
[Section 5.9] The visualization in Section 5.9 uses a UCR univariate dataset without identifying which dataset is used or explaining why a multivariate model is evaluated on a univariate sequence.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity in the derivation; the only reduction-like vulnerability (Eq. 12) is a non-circular algebraic flaw.

full rationale

I walked the claimed chain: HP-filter decomposition (Eq. 1), patch attention views (Eqs. 2-7), fusion (Eqs. 8-9), KL contrastive loss (Eqs. 10-12), anomaly score (Eq. 13), and benchmark F1 (Table 4). Each component is introduced from an external or architectural choice rather than fitted to the target F1 values; the reported numbers are empirical measurements, not outputs of a fitted constant, and no load-bearing argument rests on a self-citation. The one place where an equation reduces identically is Eq. 12: because Eq. 10 equals KL(A||B)+KL(B||A) and Eq. 11 equals the same value with arguments swapped, L_intra = L_inter for every input and L = 0. That is a serious internal-consistency/correctness problem (the written training objective cannot train the encoder), but it is not a circular derivation in which a prediction is equivalent to its input; it is a self-canceling objective. I therefore classify it outside the circularity score. Threshold selection and the unproven KL premise also raise soundness concerns, but none of them makes the central F1 claim equivalent to the paper's assumptions by construction.

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

No new physical or mathematical entities are introduced. The only novel objects are model components such as inter-patch and intra-patch views and the asynchronous loss, which are architecture constructs rather than postulated entities requiring independent evidence. The central claim rests primarily on hand-chosen per-dataset parameters and on the unproven assumption that cross-view KL divergence marks anomalies.

free parameters (5)

HP filter smoothing parameter alpha = not reported
Controls the trend/cycle split in Equation (1); chosen by hand with no value or selection rule given, and it affects all downstream representations.
Patch size set per dataset = MSL [3,5]; SMAP [3,5,7]; PSM [1,3,5]; SMD [5,7]; SWaT [3,5,7]
Table 3 lists different patch sizes per dataset, and Section 5.7 shows F1 varies with patch size; the choice is per-dataset and hand-tuned.
Window size per dataset = MSL 90; SMAP 105; PSM 60; SMD 105; SWaT 105
Table 3; Section 5.7 shows window size materially changes F1, so it is a hand-tuned choice affecting results.
Anomaly threshold rho = not reported
Equation (14) thresholds the KL score, but no value or selection procedure is given, and F1 depends directly on it.
Number of transformer encoder layers = contradictory: 2 in implementation, 3 in sensitivity analysis
Section 5.4 states two encoder layers; Section 5.8 concludes three layers are optimal. This ambiguity is a free choice that changes model capacity and results.

assumptions (4)

domain assumption HP filter decomposes each window into additive trend and cyclical components that are sufficient for anomaly detection
Section 4.2 introduces Equation (1) and asserts anomalies manifest differently across components, but no evidence is given that this decomposition is adequate for these datasets.
domain assumption KL divergence between intra-patch and inter-patch attention matrices is a monotone anomaly score
Section 4.6 uses this divergence as the anomaly score; it is the load-bearing premise that normal data yields consistent views and anomalies do not.
domain assumption Patch segmentation assumes T = P times N with no padding or residual handling
Section 4.4 uses T = P times N; there is no discussion of windows not divisible by a patch size or of discarded timestamps.
domain assumption Stop-gradient prevents collapse in this attention-matrix setting
Section 4.5 cites SimSiam; the transfer of the stop-gradient result to attention matrices and a KL loss is assumed rather than proven here.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Decomposition-based multi-scale transformer framework for time series anomaly detection." pith.science (2026). https://pith.science/paper/5KOUJEZ4

@misc{pith2026250414206,
  author       = {Pith},
  title        = {Pith review of: Decomposition-based multi-scale transformer framework for time series anomaly detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5KOUJEZ4}},
  note         = {Machine review of arXiv:2504.14206}
}

read the original abstract

Time series anomaly detection is crucial for maintaining stable systems. Existing methods face two main challenges. First, it is difficult to directly model the dependencies of diverse and complex patterns within the sequences. Second, many methods that optimize parameters using mean squared error struggle with noise in the time series, leading to performance deterioration. To address these challenges, we propose a transformer-based framework built on decomposition (TransDe) for multivariate time series anomaly detection. The key idea is to combine the strengths of time series decomposition and transformers to effectively learn the complex patterns in normal time series data. A multi-scale patch-based transformer architecture is proposed to exploit the representative dependencies of each decomposed component of the time series. Furthermore, a contrastive learn paradigm based on patch operation is proposed, which leverages KL divergence to align the positive pairs, namely the pure representations of normal patterns between different patch-level views. A novel asynchronous loss function with a stop-gradient strategy is further introduced to enhance the performance of TransDe effectively. It can avoid time-consuming and labor-intensive computation costs in the optimization process. Extensive experiments on five public datasets are conducted and TransDe shows superiority compared with twelve baselines in terms of F1 score. Our code is available at https://github.com/shaieesss/TransDe.

Figures

Figures reproduced from arXiv: 2504.14206 by the authors.

**Figure 2.** Attention mechanism for inter-patch and intra-patch views [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗

**Figure 3.** A simple example of dimension expansion for different patch views [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: The ablation experiments of normalization operation [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]

**Figure 5.** Figure 5: The ablation experiments of contrastive paradigms [PITH_FULL_IMAGE:figures/full_fig_p023_5.png]

**Figure 6.** Figure 6: The model performance of experiments on window size [PITH_FULL_IMAGE:figures/full_fig_p024_6.png]

**Figure 7.** Figure 7: The observations of hyperparameter sensitivity experiments on five datasets [PITH_FULL_IMAGE:figures/full_fig_p025_7.png]

**Figure 8.** Figure 8: Visualization The figure shows that anomalies are present in the original data, specifically between timestamps 500 to 800 and 2300 to 2600, and these are successfully identified in the predicted labels. This highlights the effectiveness of TransDe in detecting anoma…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 11 canonical work pages

[1]

Abdulaal, A., Liu, Z., and Lancewicki, T. (2021). Practical approach to asynchronous multivariate time series anomaly detection and localization. In Proceedings of the 27th acm sigkdd conference on knowledge discovery & data mining , pages 2485–2494. Bai, N., Wang, X., Han, R., Wang, Q., and Liu, Z. (2023). Paformer: anomaly detection of time series with ...

work page 2021
[36]

and Affenzeller, M

Muhr, D. and Affenzeller, M. (2022). Outlier/anomaly detection of univari- ate time series: a dataset collection and benchmark. In Lecture notes in computer science, pages 163 –

work page 2022
[72]

Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., and Pei, D. (2019a). Robust anomaly detection for multivariate time series through stochastic recur- rent neural network. In Proceedings of the 25th acm sigkdd international conferencce on knowledge discovery and data mining , pages 2828–2837. Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., and Pei, D. (2019b). Ro...

work page 2019
[127]

H., Wu, Q

Torabi, M., Rasouli, A. H., Wu, Q. M. J., Cao, W., and Pourpanah, F. (2024). Self-supervised adversarial adaptation network for breast cancer detection. Engineering applications of artificial intelligence , 133(E). 31 Tuli, S., Casale, G., and Jennings, N. R. (2022). Tranad: deep transformer networks for anomaly detection in multivariate time series data....

work page 2024
[151]

Yang, Y., Zhang, C., Zhou, T., Wen, Q., and Sun, L. (2023). Dcdetector: dual attention contrastive representation learning for time series anomaly detection. In Proceedings of the 29th ACM SIGKDD conference on knowl- edge discovery and data mining , pages 3033–3045. Yu, J., Gao, X., Li, B., Zhai, F., Lu, J., Xue, B., Fu, S., and Xiao, C. (2024). A filter-...

work page 2023
[169]

Oord, A. v. d., Li, Y., and Vinyals, O. (2018). Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748 . Paparrizos, J., Kang, Y., Boniol, P., Tsay, R. S., Palpanas, T., and Franklin, M. J. (2022). Tsb-uad: an end-to-end benchmark suite for univariate time-series anomaly detection. Proceedings of the VLDB Endowment , 1...

arXiv 2018
[238]

X., and Lin, D

Wu, Z., Xiong, Y., Yu, S. X., and Lin, D. (2018). Unsupervised feature learning via non-parametric instance discrimination. In 2018 IEEE/CVF conference on computer vision and pattern recognition , pages 3733–3742. Xu, H., Wang, Y., Pang, G., Jian, S., Liu, N., and Wang, Y. (2023a). Rosas: Deep semi-supervised anomaly detection with contamination-resilient...

arXiv 2018
[258]

H., Buchatskaya, E., Doersch, C., Pires, B

Grill, J.-B., Strub, F., Altche, F., Tallec, C., Richemond, P. H., Buchatskaya, E., Doersch, C., Pires, B. A., Guo, Z. D., Azar, M. G., Piot, B., Kavukcuoglu, K., Munos, R., and Valko, M. (2020). Bootstrap your own latent a new approach to self-supervised learning. In Advances in neural information processing systems, volume 2020-December. 28 He, K., Fan,...

work page 2020

Show all 13 references

[271]

29 Li, Z., Zhao, Y., Han, J., Su, Y., Jiao, R., Wen, X., and Pei, D. (2021). Multivariate time series anomaly detection and interpretation using hier- archical inter-metric and temporal embedding. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data m...

2021
[290]

Kong, S., Ai, J., Lu, M., and Gong, Y. (2024). Grand: Gan-based soft- ware runtime anomaly detection method using trace information. Neural networks, 169:365–377. Li, Y., Peng, X., Wu, Z., Yang, F., He, X., and Li, Z. (2023). M3gan: A masking strategy with a mutable filter for...

2024
[493]

Yu, Q., Jibin, L., and Jiang, L. (2016). An improved arima-based traffic anomaly detection algorithm for wireless sensor networks. International journal of distributed sensor networks . Zhang, C., Zhou, T., Wen, Q., and Sun, L. (2022). Tfad: A decomposition time series anomaly...

2016
[657]

Gao, X., Yu, J., Zha, S., Fu, S., Xue, B., Ye, P., Huang, Z., and Zhang, G. (2022). An ensemble-based outlier detection method for clustered and local outliers with differential potential spread loss. Knowledge-based systems,

2022
[4306]

Mathur, A. P. and Tippenhauer, N. O. (2016). Swat: A water treatment testbed for research and training on ics security. In 2016 international workshop on cyber-physical systems for smart water networks , pages 31–

2016

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Abdulaal, A., Liu, Z., and Lancewicki, T. (2021). Practical approach to asynchronous multivariate time series anomaly detection and localization. In Proceedings of the 27th acm sigkdd conference on knowledge discovery & data mining , pages 2485–2494. Bai, N., Wang, X., Han, R., Wang, Q., and Liu, Z. (2023). Paformer: anomaly detection of time series with ...

work page 2021

[2] [36]

and Affenzeller, M

Muhr, D. and Affenzeller, M. (2022). Outlier/anomaly detection of univari- ate time series: a dataset collection and benchmark. In Lecture notes in computer science, pages 163 –

work page 2022

[3] [72]

Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., and Pei, D. (2019a). Robust anomaly detection for multivariate time series through stochastic recur- rent neural network. In Proceedings of the 25th acm sigkdd international conferencce on knowledge discovery and data mining , pages 2828–2837. Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., and Pei, D. (2019b). Ro...

work page 2019

[4] [127]

H., Wu, Q

Torabi, M., Rasouli, A. H., Wu, Q. M. J., Cao, W., and Pourpanah, F. (2024). Self-supervised adversarial adaptation network for breast cancer detection. Engineering applications of artificial intelligence , 133(E). 31 Tuli, S., Casale, G., and Jennings, N. R. (2022). Tranad: deep transformer networks for anomaly detection in multivariate time series data....

work page 2024

[5] [151]

Yang, Y., Zhang, C., Zhou, T., Wen, Q., and Sun, L. (2023). Dcdetector: dual attention contrastive representation learning for time series anomaly detection. In Proceedings of the 29th ACM SIGKDD conference on knowl- edge discovery and data mining , pages 3033–3045. Yu, J., Gao, X., Li, B., Zhai, F., Lu, J., Xue, B., Fu, S., and Xiao, C. (2024). A filter-...

work page 2023

[6] [169]

Oord, A. v. d., Li, Y., and Vinyals, O. (2018). Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748 . Paparrizos, J., Kang, Y., Boniol, P., Tsay, R. S., Palpanas, T., and Franklin, M. J. (2022). Tsb-uad: an end-to-end benchmark suite for univariate time-series anomaly detection. Proceedings of the VLDB Endowment , 1...

arXiv 2018

[7] [238]

X., and Lin, D

Wu, Z., Xiong, Y., Yu, S. X., and Lin, D. (2018). Unsupervised feature learning via non-parametric instance discrimination. In 2018 IEEE/CVF conference on computer vision and pattern recognition , pages 3733–3742. Xu, H., Wang, Y., Pang, G., Jian, S., Liu, N., and Wang, Y. (2023a). Rosas: Deep semi-supervised anomaly detection with contamination-resilient...

arXiv 2018

[8] [258]

H., Buchatskaya, E., Doersch, C., Pires, B

Grill, J.-B., Strub, F., Altche, F., Tallec, C., Richemond, P. H., Buchatskaya, E., Doersch, C., Pires, B. A., Guo, Z. D., Azar, M. G., Piot, B., Kavukcuoglu, K., Munos, R., and Valko, M. (2020). Bootstrap your own latent a new approach to self-supervised learning. In Advances in neural information processing systems, volume 2020-December. 28 He, K., Fan,...

work page 2020