Pith. sign in

REVIEW 3 major objections 6 minor 22 references

GDformer: Going Beyond Subsequence Isolation for Multivariate Time Series Anomaly Detection

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Replacing subsequence-isolated self-attention with a global dictionary and prototype scoring yields a unified anomaly detection criterion that beats prior methods on all evaluated benchmarks.

desk verdict Replacing self-attention with a learned global dictionary and prototype-similarity scoring is a real improvement, but the per-dataset tuned threshold muddies the 'unsupervised SOTA' claim and needs clearer disclosure. read the letter →

arxiv 2501.18196 v3 pith:7CRXFINJ submitted 2025-01-30 cs.LG

Qingxiang Liu , Xiaoliang Luo , Chenghao Liu , Sheng Sun , Di Yao , Lvchun Wang , Wei Yu , Yuxuan Liang This is my paper · ORCID
classification cs.LG
keywords multivariatetimeseriesanomalydetectionunsupervisedlearningtransformercross-attentionglobaldictionaryprototypessubsequenceisolation
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

Existing transformer-based unsupervised anomaly detectors score points by local self-attention inside short subsequences and then stitch subsequence-level scores into a global criterion, which the paper argues causes false positives and false negatives. GDformer instead lets every time point attend to a small global dictionary of learned keys and values shared across the entire series, so all points are scored against the same normal reference. Prototypes capture the typical distribution of these point-to-dictionary attention weights, and a point is anomalous when its attention pattern is unlike the prototypes. The paper reports the best F1 scores on the MSL, SMAP, SWaT, and PSM benchmarks (plus two additional sets in the appendix), and shows the learned dictionary transfers across datasets with little loss. It also states that the theoretical basis for the key-value assignment is left to future work.

What carries the argument

The load-bearing mechanism is the dictionary-based cross-attention map: instead of attending to other tokens in a subsequence, each temporal token attends to a global dictionary of $N$ key-value pairs that is shared by all normal points across the entire series. The resulting row of the attention map, $M^h_l$, is a correlation distribution over the dictionary, and the prototypes $E_l \in \mathbb{R}^{P \times N}$ (softmax-normalized) serve as learned templates for what a normal point's correlation distribution looks like. The similarity score $S^h_l = M^h_l \operatorname{Softmax}(E_l)^\top$ converts that distribution into a vector of prototype-matching strengths, whose row-sum is the point's normality score. This object carries the argument because it makes every point's score comparable against a fixed, series-level reference, which is exactly what subsequence-isolated attention cannot provide.

What would settle it

Run GDformer on a multivariate series that contains a normal operating-mode shift with no anomaly, and check whether points in the new mode have lower similarity to the prototypes than points in the old mode; if they do, the series-level similarity criterion produces false positives that a subsequence-based method would not.

Watch

Extended reading notes

Core claim

The central claim is that a single series-level detection criterion can be derived directly from the correlation weights between each time point and a global dictionary of normal patterns, without any subsequence-level aggregation. In each layer, the dictionary-based cross-attention produces a matrix $M^h_l = \operatorname{Softmax}(Q^h_l K^{h\top}_l / \sqrt{D_h})$ whose rows describe how each point's query distributes attention over the $N$ learned key vectors; because the dictionary is shared across the whole series, these rows are directly comparable. A set of $P$ learned prototypes per layer, normalized by softmax, captures the typical row shapes of normal points, and the similarity $S^h_l = M^h_l \operatorname{Softmax}(E_l)^\top$ measures how close a point's correlation pattern is to the prototypes. The training loss maximizes this similarity for normal points while reconstructing the input, and the inference criterion is the softmax of the negative summed similarity, thresholded by a per-dataset $\delta$. The paper reports state-of-the-art F1 scores on all four standard benchmarks and shows the dictionary and prototypes transfer across datasets, while noting that the theoretical basis for the key-value assignment is left to future work.

Load-bearing premise

Anomalies of every type will always look less similar to the learned normal prototypes than normal points do, so a single per-dataset threshold can separate them without ever being shown an anomaly.

Editorial extensions

If this is right

  • The unified series-level criterion removes the need to reconcile per-subsequence scores, which the paper shows suppresses the false positives and false negatives that arise when subsequences are heterogeneous.
  • The global dictionary and prototypes transfer across datasets with only a small F1 drop, suggesting that normal temporal patterns are partly shared across domains and that a single pretrained detector could be adapted to new systems.
  • The $O(TN)$ cross-attention replaces $O(T^2)$ self-attention, cutting training time by roughly 89% versus AnomalyTrans and 95% versus DCdetector while using far less GPU memory.
  • The similarity-based criterion, rather than reconstruction error, drives most of the performance gain: ablations show it improves average F1 by about 9 points over reconstruction-based scoring.

Reading between the lines

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

  • The dictionary keys could be inspected as learned prototypical temporal shapes (trends, spikes, periodic components); if they are interpretable, the same architecture would double as a diagnostic tool that explains why a point is anomalous, not just that it is.
  • The per-dataset threshold $\delta$ is set using the known anomaly proportion, so the method as presented is semi-supervised at the decision boundary; a fully unsupervised variant would need a label-free thresholding rule, and its performance would be the real test of the 'unified criterion' claim.
  • The cross-dataset transfer results suggest a pretrained dictionary could be fine-tuned on new sensors with very little data, which is the direction the paper names as future work on foundation models for anomaly detection.
  • Because the attention complexity is $O(TN)$ and the dictionary is small, the method could be applied to much longer windows or streaming settings where full self-attention is infeasible, making the global-context benefit accessible in practice.
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

3 major / 6 minor

Summary. The paper proposes GDformer, a Transformer variant for unsupervised multivariate time series anomaly detection. Instead of computing self-attention over isolated subsequences, GDformer uses a small learnable global dictionary of Key/Value vectors in each layer and computes cross-attention between the input tokens and this dictionary, giving O(TN) attention cost. The cross-attention rows are compared against a set of learned prototypes via a similarity score, and the negative similarity is used as the anomaly score. Training combines a reconstruction loss and a similarity loss that encourages normal points to have high prototype similarity. The authors report state-of-the-art F1 scores on MSL, SMAP, SWaT, and PSM, along with ablations, efficiency comparisons, transferability experiments, and additional evaluations on two further datasets. The code is stated to be available.

Significance. If the reported results hold under a fair, label-free thresholding protocol, the paper would make a useful contribution: the dictionary-based cross-attention is a clean and computationally efficient mechanism, the ablation study (§4.2.2) supports the contribution of each component, and the transferability experiment (§4.1, Table 2) is an interesting demonstration that the learned dictionary/prototypes generalize across datasets. The complexity analysis in Appendix C is careful. The central limitation is that the headline 'unified series-level criterion' depends on a dataset-specific threshold δ that appears to be tuned on labels, which compromises the unsupervised claim and the fairness of the SOTA comparison.

major comments (3)
  1. [§3.2, Eq. (6); Table 5; Appendix A] The detection rule in Eq. (6) depends on a threshold δ, and Table 5 reports dataset-specific values (0.8, 0.7, 0.5, 0.6) described as 'top δ% anomaly score is termed as the detection criterion.' The paper does not state how δ is selected. If δ is tuned using validation/test labels to maximize F1, the method is not purely unsupervised and the 'unified series-level criterion' is not unified across datasets. Because AnomalyScore in Eq. (5) is a monotone softmax of the negative learned similarity, thresholding at top δ% is equivalent to flagging the δ fraction of test points with the lowest prototype similarity. I ask the authors to (a) specify exactly how δ is chosen, and (b) report results under a label-free thresholding protocol—for example, an extreme-value or fixed-quantile threshold fitted only on normal training scores—and show that the SOTA margins over AnomalyTransformer and DCdetector survive. The transferability results in Table 2 are also affected if δ is re-tuned per target dataset.
  2. [Abstract; §4.1; Table 1; Appendix D.3] The abstract claims 'consistently achieves state-of-the-art unsupervised anomaly detection performance on five real-world benchmark datasets,' but Section 4 states 'We evaluate on 4 real-world benchmark datasets' and Table 1 reports results for only MSL, SMAP, SWaT, and PSM. Two additional datasets (NIPS_TS_GECCO and ASD) appear in Table 7, but they are not part of the main comparison and no dataset statistics or hyperparameter settings are given for them. The central claim of consistent SOTA across five benchmarks is therefore not supported by the paper as written. The authors should either move the additional datasets into the main results with full experimental details or revise the claim to 'four datasets in the main comparison, with additional results in the appendix.'
  3. [§3.2, Eq. (4); §3.1.2] Equation (4) trains the prototypes by maximizing the same similarity that Eq. (5)–(6) later use for detection: Ls sums the similarity between cross-attention rows and prototypes on normal points, so 'normal points have high prototype similarity' is true by construction. This is not inherently a flaw—one-class methods such as Deep-SVDD also learn a score on normal data—but the wording in §3.1.2 that the training 'naturally leads to a representation-wise similarity-based detection criterion' overstates the derivation. The empirical question is whether the learned similarity separates anomalies before thresholding. To decouple representation quality from threshold tuning, please add a threshold-free evaluation of AnomalyScore, such as AUROC or PR-AUC computed on the test set without applying Eq. (6).
minor comments (6)
  1. [§3.1.1] The symbol \tilde{X} is used for both the masked, instance-normalized input and the embedded input after the linear projection; please use a distinct symbol for the embedding (e.g., X0).
  2. [Figure 1 caption] The caption contains the typo 'DCdctector'; it should read 'DCdetector.'
  3. [§2] In the related-work discussion, 'LATM-V AE' appears to be a typo for 'LSTM-VAE.'
  4. [Appendix D.1, Figure 6] The text states that '[1,3] may be an optimal range' for λ, but Figure 6(a) shows substantial F1 variation without error bars; please report the variance or avoid a quantitative range claim based on single-run curves.
  5. [Table 6] The EH-GAM-EGAN row reports F1 scores of 9.10 for SMAP and 8.63 for SWaT, which are far below typical values; please verify that the baseline was evaluated under the same protocol and clarify whether these numbers are cited or rerun.
  6. [§4.1] The sentence 'We directly cite the results from Yang et al. (2023b) if applicable' does not specify which of the 19 baselines were rerun and which were taken from the literature; please provide this information for reproducibility and fairness.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the prototype-similarity score is the training objective by design, and the dataset-specific threshold is a standard operating-point choice, not a derivation-chain reduction.

full rationale

The central derivation is self-contained. GDformer's anomaly score (Eq. 5) is a monotone function of the negative prototype-similarity S (Eq. 3), and the similarity loss Ls in Eq. (4) explicitly maximizes that same similarity on the training points. This makes "normal training points have high similarity to prototypes" true by construction, but that is a one-class training objective, not a circular prediction: the paper's empirical claim is that held-out abnormal points have low similarity, which is an assumption tested against external benchmarks (MSL, SMAP, SWaT, PSM, NIPS_TS_GECCO, ASD). No equation in the paper derives the anomaly labels from the training labels; Eq. (6) only thresholds the learned score. The per-dataset threshold delta (Table 5, Fig. 6d) is a standard operating-point selection; if it is tuned on validation labels it makes the reported F1 partly a model-selection result, but it does not reduce the learned representation to the labels or make the score itself circular. The only self-citation is Liang et al. (2024) in Sec. 3.1.2, used to motivate foundation-model transferability; it is not load-bearing for the detection criterion. The Limitations section explicitly defers theoretical analysis of the key-value pairs, a missing proof rather than a circular step. Overall, no prediction in the paper reduces by construction to its inputs; score 2 reflects only the minor, non-load-bearing self-citation.

Assumptions & free parameters 4 free parameters · 3 assumptions · 2 invented entities

The central claim depends on four per-dataset tuned hyperparameters (λ, P, N, δ), which together set the detection boundary. The method also assumes clean training data and a monotone relationship between prototype similarity and abnormality, with no theoretical justification. The global dictionary and prototypes are learned components supported by ablation and transfer evidence.

free parameters (4)
  • λ (loss trade-off) = 3, 2, 2, 1 for MSL, SMAP, SWaT, PSM
    Tuned per dataset (Table 5); balances reconstruction loss and prototype similarity loss.
  • P (number of prototypes) = 12, 12, 8, 10 for MSL, SMAP, SWaT, PSM
    Tuned per dataset; affects compactness of the decision boundary.
  • N (dictionary size) = 16, 6, 8, 10 for MSL, SMAP, SWaT, PSM
    Tuned per dataset; determines capacity of the global dictionary.
  • δ (detection threshold) = 0.8, 0.7, 0.5, 0.6 for MSL, SMAP, SWaT, PSM
    Tuned per dataset (Table 5); directly sets the top-percentile anomaly criterion in Eq. (6), effectively fitting the decision boundary to each benchmark.
assumptions (3)
  • domain assumption Training subsequences are anomaly-free and representative of normal behavior
    The method assumes the learned global dictionary is not contaminated by anomalies; the paper uses standard benchmark splits but does not explicitly state the training data are clean.
  • domain assumption Dot-product similarity between cross-attention rows and softmaxed prototypes is a monotone proxy for abnormality
    Eq. (3) assumes lower similarity to normal prototypes implies higher anomaly likelihood for all anomaly types; the Limitations section admits no theoretical analysis is provided.
  • standard math Adam optimization converges to a useful solution for the combined reconstruction and similarity losses
    Relies on standard deep learning optimization practice; no convergence guarantee is given.
invented entities (2)
  • Global dictionary (learnable Key and Value vectors) independent evidence
    purpose: Provides a compact set of global normal representations for cross-attention, replacing subsequence-specific self-attention.
    Ablation (Table 3) and cross-dataset transfer experiments (Table 2) provide external falsifiable support beyond the in-sample fit.
  • Prototypes (learned normal attention-weight distributions) independent evidence
    purpose: Captures the distribution of point-to-dictionary correlation weights for normal points and forms the similarity-based detection criterion.
    Ablations and transfer results support their role; they are trained parameters rather than a priori entities.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GDformer: Going Beyond Subsequence Isolation for Multivariate Time Series Anomaly Detection." pith.science (2026). https://pith.science/paper/7CRXFINJ

@misc{pith2026250118196,
  author       = {Pith},
  title        = {Pith review of: GDformer: Going Beyond Subsequence Isolation for Multivariate Time Series Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7CRXFINJ}},
  note         = {Machine review of arXiv:2501.18196}
}
read the original abstract

Unsupervised anomaly detection of multivariate time series is a challenging task, given the requirements of deriving a compact detection criterion without accessing the anomaly points. The existing methods are mainly based on reconstruction error or association divergence, which are both confined to isolated subsequences with limited horizons, hardly promising unified series-level criterion. In this paper, we propose the Global Dictionary-enhanced Transformer (GDformer) with a renovated dictionary-based cross attention mechanism to cultivate the global representations shared by all normal points in the entire series. Accordingly, the cross-attention maps reflect the correlation weights between the point and global representations, which naturally leads to the representation-wise similarity-based detection criterion. To foster more compact detection boundary, prototypes are introduced to capture the distribution of normal point-global correlation weights. GDformer consistently achieves state-of-the-art unsupervised anomaly detection performance on five real-world benchmark datasets. Further experiments validate the global dictionary has great transferability among various datasets.

Figures

Figures reproduced from arXiv: 2501.18196 by the authors.

Figure 1
Figure 1. How to derive the detection criterion. Left: AnomalyTrans and DCdctector learn intra￾subsequence point-wise association and derive the detection criterion by combining subsequence￾level anomaly scores. Right: Our proposal cultivates global normal representations manifested as dictionary and prototypes for evaluating similarity discrepancy to provide series-level criterion. association-based. In the first ones, the r… view at source ↗
Figure 2
Figure 2. Anomaly scores v.s. detection criterion for [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The framework of GDformer (left). In Dictionary-Based Cross Attention (right), the global dictionary of Key and Value vectors (in cross-attention module) learns global representations shared by all normal points in the entire series. The prototypes (in similarity evaluation module) capture the normal distribution of cross-attention weights. 3 Methodology Suppose there are d sensors or machines in an industrial syste… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Model efficiency comparison. Larger bubble size indicates higher memory requirements. We compare the model efficiency in terms of detection accuracy, training time, and memory footprint of the following methods: Anomaly￾Trans, DCdetector, and GDformer. As shown in [PI…
Figure 5
Figure 5. Figure 5: Detection results visualization of AnomalyTrans, DCdetector, and [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Parameter sensitivity analysis of (a) loss weight [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: The showcase of prototypes (the second row) and attention maps on normal (the first row) [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 15 canonical work pages

  1. [5]

    AR: the abnormal proportion of the whole dataset

    Table 5: Dataset details and optimal hyperparameter settings. AR: the abnormal proportion of the whole dataset. Dataset d T #Training #Validation #Test AR λ P N δ MSL 55 100 46,653 11,664 73,729 0.105 3 12 16 0.8 SMAP 25 100 108,146 27,037 427,617 0.128 2 12 6 0.7 SWaT 51 100 396,000 99,000 449,919 0.121 2 8 8 0.5 PSM 25 100 105,984 26,497 87,841 0.278 1 ...

    work page 2022

  2. [6]

    Kingma, D

    URL https://openreview.net/forum?id=cGDAkQo1C0p. Kingma, D. P. and Ba, J. Adam: A method for stochastic optimization. In Bengio, Y . and LeCun, Y . (eds.),3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings,

    work page 2015

  3. [7]

    Specifically, compared with DCdetector, GDformer can promise 57.32% F1-score improvements on NIPS_TS_GECCO dataset

    GDformer outperforms the two baselines on both datasets. Specifically, compared with DCdetector, GDformer can promise 57.32% F1-score improvements on NIPS_TS_GECCO dataset. 15 1 2 3 4 5 0.840.860.880.900.920.940.960.98F1/s955 MSL SMAP SWaT PSM( a)6 8 1 01 21 60.700.750.800.850.900.95F1P (b)6 8 1 01 21 60.750.800.850.900.95F1N (c)0 .50 .60 .70 .80 .90.920 ...

    work page 2022

  4. [11]

    doi: 10.1109/TC.2021. 3065073. Tang, J., Chen, Z., Fu, A. W.-C., and Cheung, D. W. Enhancing effectiveness of outlier detections for low density patterns. In Advances in Knowledge Discovery and Data Mining: 6th Pacific-Asia Conference, PAKDD 2002 Taipei, Taiwan, May 6–8, 2002 Proceedings 6, pp. 535–548. Springer,

    work page doi:10.1109/tc.2021 2021

  5. [14]

    SGDP: A Stream-Graph Neural Network Based Data Prefetcher

    Yang, Y ., Li, R., Shi, Q., Li, X., Hu, G., Li, X., and Yuan, M. Sgdp: A stream-graph neural network based data prefetcher, 2023a. URL https://arxiv.org/abs/2304.03864. 12 Yang, Y ., Zhang, C., Zhou, T., Wen, Q., and Sun, L. Dcdetector: Dual attention contrastive repre- sentation learning for time series anomaly detection. In Proc. 29th ACM SIGKDD Interna...

    work page Pith review arXiv 2023

  6. [16]

    URL http://dx.doi.org/10.1145/3511808.3557470

    doi: 10.1145/3511808.3557470. URL http://dx.doi.org/10.1145/3511808.3557470. Zhao, H., Wang, Y ., Duan, J., Huang, C., Cao, D., Tong, Y ., Xu, B., Bai, J., Tong, J., and Zhang, Q. Multivariate time-series anomaly detection via graph attention network,

    arXiv

  7. [18]

    13 A Implementation Details Datasets. We evaluate the anomaly detection performance on 4 real-world datasets: • MSL (Mars Science Laboratory dataset) is collected by NASA with 55 dimensions and shows the condition of the sensors and actuator data from the Mars rover Hundman et al. (2018a). • SMAP (Soil Moisture Active Passive dataset) is also collected fr...

    work page 2018

  8. [20]

    The training process is continued for 10 epochs with the batch size of

    We employ the ADAM Kingma & Ba (2015) with an initial learning rate of 10−4 to optimize model parameters. The training process is continued for 10 epochs with the batch size of

    work page 2015

Show all 22 references
  1. [64]

    (2019) with a single NVIDIA GeForce RTX 3090 24GB GPU

    All experiments are implemented in PyTorch Paszke et al. (2019) with a single NVIDIA GeForce RTX 3090 24GB GPU. B Process of Dictionary-based Cross Attention Algorithm 1: Dictionary-based Cross Attention Mechanism Input: Xl−1 ∈ RT ×D; Kh l ∈ RN ×Dh, V h l ∈ RN ×Dh (h ∈ [1, H])...

    2019

  2. [1970]

    M., Kriegel, H.-P., Ng, R

    Breunig, M. M., Kriegel, H.-P., Ng, R. T., and Sander, J. Lof: identifying density-based local outliers. In Proceedings of the 2000 ACM SIGMOD international conference on Management of data, pp. 93–104,

    2000

  3. [2000]

    V ., Xue, H., and Salim, F

    Deldari, S., Smith, D. V ., Xue, H., and Salim, F. D. Time series change point detection with self-supervised contrastive predictive coding. In Proceedings of the Web Conference 2021, pp. 3124–3135,

    2021

  4. [2004]

    He, Y ., Chen, X., Miao, D., Zhang, H., Qin, X., Du, S., and Lu, P

    doi: 10.1109/ ISIT.2004.1365067. He, Y ., Chen, X., Miao, D., Zhang, H., Qin, X., Du, S., and Lu, P. Graph-enhanced anomaly detection framework in multivariate time series using graph attention and enhanced generative adversarial networks. Expert Systems with Applications, 271:126667,

    2004 arXiv

  5. [2008]

    Mathur, A. P. and Tippenhauer, N. O. Swat: a water treatment testbed for research and training on ics security. In 2016 International Workshop on Cyber-physical Systems for Smart Water Networks (CySWater), pp. 31–36,

    2016

  6. [2014]

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A

    doi: 10.1109/TIT.2014.2320500. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L. u., and Polosukhin, I. Attention is all you need. In Guyon, I., Luxburg, U. V ., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., and Garnett, R. (eds....

    2014

  7. [2016]

    Paparrizos, J., Boniol, P., Palpanas, T., Tsay, R

    doi: 10.1109/CySWater.2016.7469060. Paparrizos, J., Boniol, P., Palpanas, T., Tsay, R. S., Elmore, A., and Franklin, M. J. V olume under the surface: a new accuracy evaluation measure for time-series anomaly detection. Proceedings of the VLDB Endowment, 15(11):2774–2787,

    2016

  8. [2017]

    Wen, Q., Yang, L., Zhou, T., and Sun, L

    URL https://proceedings.neurips.cc/ paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf. Wen, Q., Yang, L., Zhou, T., and Sun, L. Robust time series analysis and applications: An industrial perspective. In Proceedings of the 28th ACM SIGKDD Conference on Kno...

    2017

  9. [2018]

    Zhang, C., Zhou, T., Wen, Q., and Sun, L

    URL https://arxiv.org/abs/1811.08055. Zhang, C., Zhou, T., Wen, Q., and Sun, L. Tfad: A decomposition time series anomaly detection architecture with time-frequency analysis. In Proceedings of the 31st ACM International Con- ference on Information & Knowledge Management, pp. 2...

    arXiv

  10. [2019]

    Perslev, M., Jensen, M., Darkner, S., Jennum, P

    URL https://proceedings.neurips.cc/paper_files/paper/2019/ file/bdbca288fee7f92f2bfa9f7012727740-Paper.pdf. Perslev, M., Jensen, M., Darkner, S., Jennum, P. J., and Igel, C. U-time: A fully convolutional network for time series segmentation applied to sleep staging. Advances i...

    2019

  11. [2020]

    Zhou, B., Liu, S., Hooi, B., Cheng, X., and Ye, J

    URL https: //arxiv.org/abs/2009.02040. Zhou, B., Liu, S., Hooi, B., Cheng, X., and Ye, J. Beatgan: Anomalous rhythm detection using adversarially generated time series. In IJCAI, volume 2019, pp. 4433–4439,

    2009 arXiv

  12. [2021]

    ISBN 9781450383325

    Association for Computing Machinery. ISBN 9781450383325. doi: 10.1145/3447548.3467174. URL https://doi.org/10.1145/3447548.3467174. Adams, R. P. and MacKay, D. J. Bayesian online changepoint detection. arXiv preprint arXiv:0710.3742,

  13. [2022]

    Detecting space- craft anomalies using lstms and nonparametric dynamic thresholding

    Hundman, K., Constantinou, V ., Laporte, C., Colwell, I., and Soderstrom, T. Detecting space- craft anomalies using lstms and nonparametric dynamic thresholding. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , KDD ’18, pp. ...

  14. [2024]

    T., Ting, K

    Liu, F. T., Ting, K. M., and Zhou, Z.-H. Isolation forest. In 2008 eighth ieee international conference on data mining, pp. 413–422. IEEE,

    2008

Pith tools

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