arxiv: 2604.08582 · v1 · submitted 2026-03-29 · 💻 cs.LG · cs.AI

Recognition: 1 theorem link

· Lean Theorem

Multivariate Time Series Anomaly Detection via Dual-Branch Reconstruction and Autoregressive Flow-based Residual Density Estimation

Jun Liu, Jun Tang, Qinyue Tong, Ying Chen, Ziqian Lu

Authors on Pith no claims yet

Pith reviewed 2026-05-14 22:04 UTC · model grok-4.3

classification 💻 cs.LG cs.AI

keywords multivariate time seriesanomaly detectionreconstruction-based methodsautoregressive flowresidual density estimationdual-branch encoderspurious correlations

0 comments

The pith

DBR-AF decouples cross-variable and intra-variable modeling then applies autoregressive flow density estimation on residuals to detect multivariate time series anomalies more reliably.

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

Mainstream reconstruction methods for multivariate time series anomaly detection overfit to spurious cross-variable correlations and produce misleading scores by simply summing reconstruction errors. The paper introduces DBR-AF, which uses a dual-branch encoder to separate correlation learning from intra-variable statistics and an autoregressive flow module to model residual distributions via density estimation. This combination aims to distinguish genuine anomalies from merely hard-to-reconstruct normal samples. Experiments across seven benchmark datasets show state-of-the-art results, with ablations confirming that both the decoupling and the flow-based scoring are required.

Core claim

The DBR encoder decouples cross-variable correlation learning from intra-variable statistical property modeling to reduce overfitting to spurious correlations, while the AF module stacks reversible transformations to capture the complex multivariate residual distribution and applies density estimation to assign accurate anomaly scores even to normal samples with large reconstruction errors.

What carries the argument

Dual-branch reconstruction (DBR) encoder that splits cross-variable and intra-variable paths, combined with autoregressive flow (AF) module for residual density estimation via stacked reversible transformations.

If this is right

Anomaly scores become more trustworthy because density estimation on residuals separates hard normal samples from actual anomalies.
Overfitting to spurious correlations decreases when the encoder is forced to model intra-variable statistics independently.
The framework scales to real monitoring tasks in industrial control and aerospace systems where summed error scores currently fail.
Ablation results indicate that removing either the dual-branch structure or the flow module drops performance on the seven benchmarks.

Where Pith is reading between the lines

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

The same decoupling strategy could be tested on other reconstruction-based detectors to check whether it reduces spurious fits in non-time-series domains.
Density estimation on residuals might improve scoring in related tasks such as multivariate forecasting or imputation where reconstruction errors are also ambiguous.
If the reversible transformations in the flow module prove stable under distribution shift, the approach could extend to online anomaly detection on streaming data.

Load-bearing premise

Separating cross-variable correlation learning from intra-variable modeling will reduce spurious correlations without discarding information needed to detect true anomalies.

What would settle it

A dataset in which strong causal cross-variable dependencies are the primary signal of anomalies, where the dual-branch model underperforms a joint-model baseline on held-out test data.

Figures

Figures reproduced from arXiv: 2604.08582 by Jun Liu, Jun Tang, Qinyue Tong, Ying Chen, Ziqian Lu.

**Figure 1.** Figure 1: The left subplot reveals the fatal flaws of existing reconstruction-based methods: cross-variable interference induces large amplitude errors and spurious non-stationary characteristics in reconstructed signals, and anomaly scores are dominated by variables with significant reconstruction errors, thereby giving rise to false positive detections. In contrast, the right subplot demonstrates the advantages of… view at source ↗

**Figure 2.** Figure 2: Overall Framework Diagram of Our Method DBR-AF. Without loss of generality, each transformation in our autoregressive flow is represented as a Scale & Shift operation. The dashed box illustrates the distribution transformation in the autoregressive flow module. Specifically, the reconstructed residual is given by x residual = {x residual 1 , · · · , x residual T }, and each residual at time step t is indep… view at source ↗

**Figure 3.** Figure 3: Variations of F1-score with DAF. 2) Different Prior distributionint in the Autoregressive Flow Module: For the prior distribution used in the AF module, we introduce two candidate distributions for comparative performance evaluation: the diagonal Gaussian distribution and the Gaussian mixture distribution with a variable number of components. Additionally, each of these prior distributions is configured i… view at source ↗

**Figure 4.** Figure 4: AUC-ROC and F1-score distributions by number of encoder Layers, averaged over five real-world benchmark datasets. count is too large. For the F1-score metric, the impact of different encoder layer configurations on performance is relatively minor, with optimal results achieved when both branches are set to 3 layers. 2) Window Size and Encoder Dimension of Temporal Branch: figs. 5 and 6 characterize the mod… view at source ↗

**Figure 5.** Figure 5: Variations of AUC-ROC and F1-score with Window Size. 32 64 128 256 512 Encoder Dimension 55 60 65 70 75 80 AUC-ROC SMD MSL SMAP PSM SWAT 32 64 128 256 512 Encoder Dimension 88 90 92 94 96 98 F1-Score SMD MSL SMAP PSM SWAT [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

**Figure 6.** Figure 6: Variations of AUC-ROC and F1-score with Encoder Dimension of Temporal Branch Denc [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗

**Figure 7.** Figure 7: Visualization of more anomaly detection results. From top to bottom in each subplot: time series signal, reconstruction signal, anomaly score. TABLE X F1-score under Different λ Values λ MSL SMAP PSM SWaT SMD 0.4 94.14 95.98 97.75 96.34 93.10 0.6 93.85 96.38 98.12 96.89 93.98 0.8 94.42 96.57 98.77 97.68 94.02 1.0 95.59 96.31 98.88 97.66 94.80 1.2 94.87 95.76 98.45 97.21 93.75 1.4 93.92 95.42 98.03 96.75 92… view at source ↗

read the original abstract

Multivariate Time Series Anomaly Detection (MTSAD) is critical for real-world monitoring scenarios such as industrial control and aerospace systems. Mainstream reconstruction-based anomaly detection methods suffer from two key limitations: first, overfitting to spurious correlations induced by an overemphasis on cross-variable modeling; second, the generation of misleading anomaly scores by simply summing up multivariable reconstruction errors, which makes it difficult to distinguish between hard-to-reconstruct samples and genuine anomalies. To address these issues, we propose DBR-AF, a novel framework that integrates a dual-branch reconstruction (DBR) encoder and an autoregressive flow (AF) module. The DBR encoder decouples cross-variable correlation learning and intra-variable statistical property modeling to mitigate spurious correlations, while the AF module employs multiple stacked reversible transformations to model the complex multivariate residual distribution and further leverages density estimation to accurately identify normal samples with large reconstruction errors. Extensive experiments on seven benchmark datasets demonstrate that DBR-AF achieves state-of-the-art performance, with ablation studies validating the indispensability of its core components.

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.

Desk Editor's Note private letter to a colleague

DBR-AF pairs a dual-branch encoder with autoregressive flow on residuals to cut spurious correlations and improve anomaly scoring, a reasonable incremental step if the numbers check out.

read the letter

The main takeaway is that this paper targets two practical weaknesses in reconstruction-based multivariate time series anomaly detection: overfitting to cross-variable correlations and the blunt use of summed reconstruction errors for scoring. The proposed DBR-AF splits the encoder into separate branches for correlation learning and per-variable statistics, then applies an autoregressive flow to model the residual distribution via density estimation rather than simple summation. That combination is the concrete new element here. The motivation lines up with real monitoring needs in industrial and aerospace settings, and the decoupling strategy is a direct inductive bias that could plausibly limit fitting to noise. The flow module adds a standard but flexible density tool on top of the residuals, which should help flag true anomalies even when some normal samples reconstruct poorly. Experiments are described on seven benchmarks with SOTA claims and ablations that test component necessity. If the full tables show consistent gains with proper controls and error bars, this would count as a useful engineering refinement for practitioners who rely on reconstruction baselines. The soft spots are modest but worth noting. The abstract supplies no quantitative deltas or variance measures, so the size and stability of the reported improvements remain to be confirmed from the actual results. The flow itself carries a hyperparameter for the number of stacked transformations, which could affect outcomes and needs clear reporting. No internal contradictions appear in the framing, and the logic holds together without circular definitions. This paper is aimed at researchers and engineers working on multivariate time series anomaly detection who already use reconstruction methods and want better handling of dependencies. A reader looking for targeted architectural tweaks would extract value from the design choices and comparisons. It shows coherent engagement with the literature and the problem, so it deserves a serious referee to examine the implementation details, statistical support, and reproducibility.

Referee Report

0 major / 3 minor

Summary. The manuscript proposes DBR-AF for multivariate time series anomaly detection. It introduces a dual-branch reconstruction (DBR) encoder that decouples cross-variable correlation learning from intra-variable statistical property modeling to mitigate overfitting to spurious correlations. An autoregressive flow (AF) module is added to model the complex multivariate residual distribution via stacked reversible transformations and density estimation, replacing simple summation of reconstruction errors for more reliable anomaly scoring. The authors report state-of-the-art results on seven benchmark datasets and use ablation studies to demonstrate the indispensability of the DBR and AF components.

Significance. If the results hold with proper statistical support, the work provides a clear advance in reconstruction-based MTSAD by supplying a principled architectural response to two recurring limitations: over-reliance on cross-variable modeling and crude anomaly scoring. The decoupling strategy supplies a targeted inductive bias, while the AF density estimation offers a technically coherent improvement over summed-error heuristics. Ablation validation and multi-dataset evaluation are positive features that would strengthen the contribution if the quantitative evidence is robust.

minor comments (3)

[Abstract] Abstract: the claim of SOTA performance and component indispensability is stated without any numerical metrics, error bars, or dataset-specific scores; adding at least the primary F1 or AUC values would make the central claim immediately verifiable.
[Section 3.2] Section 3.2 (AF module): the description of how the stacked reversible transformations avoid post-hoc fitting issues on the same residuals is brief; a short derivation or explicit statement of the training objective would clarify the separation from the reconstruction loss.
[Section 4] Section 4 (experiments): ablation tables would benefit from reporting standard deviations across runs or statistical significance tests to substantiate the claim that each component is indispensable.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of DBR-AF and the recommendation for minor revision. The recognition that the dual-branch reconstruction provides a targeted inductive bias and that the autoregressive flow offers a coherent improvement over summed-error scoring is encouraging. We will incorporate minor revisions to strengthen statistical support for the reported results across the seven datasets.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces DBR-AF as an architectural combination of a dual-branch reconstruction encoder (decoupling cross- and intra-variable modeling) and an autoregressive flow module for residual density estimation. These are presented as independent inductive biases motivated by stated limitations of prior reconstruction-based methods. No equations, derivations, or self-citations in the provided abstract or description reduce the claimed performance gains or component indispensability to quantities defined solely by fits on the same data. Ablation studies and benchmark results serve as external validation rather than tautological re-derivations. The framework remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard assumptions in deep learning for time series plus the domain assumption that the proposed decoupling and density estimation directly solve the identified overfitting and scoring issues.

free parameters (1)

number of stacked reversible transformations in AF
Hyperparameter chosen to model complex multivariate residual distribution; value not specified in abstract.

axioms (1)

domain assumption Multivariate time series exhibit separable cross-variable correlations and intra-variable statistical properties that can be modeled independently without loss of essential information.
Invoked in the design of the DBR encoder to mitigate spurious correlations.

pith-pipeline@v0.9.0 · 5494 in / 1249 out tokens · 48868 ms · 2026-05-14T22:04:15.665898+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
dual-branch reconstruction (DBR) encoder ... autoregressive flow (AF) module ... residuals ... negative log-likelihood

Reference graph

Works this paper leans on

48 extracted references · 48 canonical work pages · 3 internal anchors

[1]

Detecting spacecraft anomalies using lstms and nonparametric dynamic thresholding,

K. Hundman, V. Constantinou, C. Laporte, I. Colwell, and T. Soderstrom, “Detecting spacecraft anomalies using lstms and nonparametric dynamic thresholding,” in Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , 2018, pp. 387–395

work page 2018
[2]

Gecco 2018 industrial chal- lenge: Monitoring of drinking-water quality,

F. Rehbach, S. Moritz, S. Chandrasekaran, M. Rebolledo, M. Friese, and T. Bartz-Beielstein, “Gecco 2018 industrial chal- lenge: Monitoring of drinking-water quality,” Accessed: Feb , vol. 19, p. 2019, 2018

work page 2018
[3]

Multivariate time series anomaly detection: Fancy algorithms and flawed evaluation methodology,

M. El Amine Sehili and Z. Zhang, “Multivariate time series anomaly detection: Fancy algorithms and flawed evaluation methodology,” in Technology Conference on Performance Eval- uation and Benchmarking . Springer, 2023, pp. 1–17

work page 2023
[4]

A survey of deep anomaly detection in multivariate time se- ries: taxonomy, applications, and directions,

F. Wang, Y. Jiang, R. Zhang, A. Wei, J. Xie, and X. Pang, “A survey of deep anomaly detection in multivariate time se- ries: taxonomy, applications, and directions,” Sensors (Basel, Switzerland), vol. 25, no. 1, p. 190, 2025

work page 2025
[5]

Anomaly transformer: Time series anomaly detection with association discrepancy,

J. Xu, H. Wu, J. Wang, and M. Long, “Anomaly transformer: Time series anomaly detection with association discrepancy,” arXiv preprint arXiv:2110.02642 , 2021

work page arXiv 2021
[6]

Tfad: A decomposition time series anomaly detection architecture with time-frequency analysis,

C. Zhang, T. Zhou, Q. Wen, and L. Sun, “Tfad: A decomposition time series anomaly detection architecture with time-frequency analysis,” in Proceedings of the 31st ACM international confer- ence on information & knowledge management , 2022, pp. 2497– 2507

work page 2022
[7]

Dcde- tector: Dual attention contrastive representation learning for time series anomaly detection,

Y. Yang, C. Zhang, T. Zhou, Q. Wen, and L. Sun, “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 , 2023, pp. 3033–3045

work page 2023
[8]

Temporal-frequency masked autoencoders for time series anomaly detection,

Y. Fang, J. Xie, Y. Zhao, L. Chen, Y. Gao, and K. Zheng, “Temporal-frequency masked autoencoders for time series anomaly detection,” in 2024 IEEE 40th International Confer- ence on Data Engineering (ICDE) . IEEE, 2024, pp. 1228–1241

work page 2024
[9]

Learn hybrid prototypes for multivariate time series anomaly detection,

K.-Y. Shen, “Learn hybrid prototypes for multivariate time series anomaly detection,” in The Thirteenth International Con- ference on Learning Representations , 2025

work page 2025
[10]

Timesnet: Temporal 2d-variation modeling for general time series analysis,

H. Wu, T. Hu, Y. Liu, H. Zhou, J. Wang, and M. Long, “Timesnet: Temporal 2d-variation modeling for general time series analysis,” arXiv preprint arXiv:2210.02186 , 2022

work page arXiv 2022
[11]

Catch: Channel-aware multivariate time series anomaly detection via frequency patching,

X. Wu, X. Qiu, Z. Li, Y. Wang, J. Hu, C. Guo, H. Xiong, and B. Yang, “Catch: Channel-aware multivariate time series anomaly detection via frequency patching,” arXiv preprint arXiv:2410.12261, 2024

work page arXiv 2024
[12]

Multivariate time series anomaly detection by capturing coarse-grained intra-and inter- variate dependencies,

Y. Xie, H. Zhang, and M. A. Babar, “Multivariate time series anomaly detection by capturing coarse-grained intra-and inter- variate dependencies,” in Proceedings of the ACM on Web Conference 2025, 2025, pp. 697–705

work page 2025
[13]

Variational inference with nor- malizing flows,

D. Rezende and S. Mohamed, “Variational inference with nor- malizing flows,” in International conference on machine learn- ing. PMLR, 2015, pp. 1530–1538

work page 2015
[14]

On the uni- versality of volume-preserving and coupling-based normalizing flows,

F. Draxler, S. Wahl, C. Schnörr, and U. Köthe, “On the uni- versality of volume-preserving and coupling-based normalizing flows,” arXiv preprint arXiv:2402.06578 , 2024

work page arXiv 2024
[15]

D. M. Hawkins, Identification of outliers . Springer, 1980, vol. 11

work page 1980
[16]

Lof: identifying density-based local outliers,

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

work page 2000
[17]

Serial order: A parallel distributed processing approach,

M. I. Jordan, “Serial order: A parallel distributed processing approach,” in Advances in psychology. Elsevier, 1997, vol. 121, pp. 471–495

work page 1997
[18]

Long short-term memory,

S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997

work page 1997
[19]

Temporal convolutional networks for action segmentation and detection,

C. Lea, M. D. Flynn, R. Vidal, A. Reiter, and G. D. Hager, “Temporal convolutional networks for action segmentation and detection,” in proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2017, pp. 156–165

work page 2017
[20]

D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “(1986) de rumelhart, ge hinton, and rj williams, learning internal repre- sentations by error propagation, parallel distributed processing: Explorations in the microstructures of cognition, vol. i, de rumelhart and jl mcclelland (eds.) cambridge, ma: Mit press, pp. 318-362,” 1988

work page 1986
[21]

Auto-Encoding Variational Bayes

D. P. Kingma and M. Welling, “Auto-encoding variational bayes,” arXiv preprint arXiv:1312.6114 , 2013

work page internal anchor Pith review Pith/arXiv arXiv 2013
[22]

A data-driven health monitoring method for satellite housekeeping data based on probabilistic clustering and dimensionality reduction,

T. Yairi, N. Takeishi, T. Oda, Y. Nakajima, N. Nishimura, and N. Takata, “A data-driven health monitoring method for satellite housekeeping data based on probabilistic clustering and dimensionality reduction,” IEEE Transactions on Aerospace and Electronic Systems , vol. 53, no. 3, pp. 1384–1401, 2017

work page 2017
[23]

Deep autoencoding gaussian mixture model for unsupervised anomaly detection,

B. Zong, Q. Song, M. R. Min, W. Cheng, C. Lumezanu, D. Cho, and H. Chen, “Deep autoencoding gaussian mixture model for unsupervised anomaly detection,” in International conference on learning representations , 2018

work page 2018
[24]

Anomaly detection using autoen- coders with nonlinear dimensionality reduction,

M. Sakurada and T. Yairi, “Anomaly detection using autoen- coders with nonlinear dimensionality reduction,” in Proceedings of the MLSDA 2014 2nd workshop on machine learning for sensory data analysis , 2014, pp. 4–11

work page 2014
[25]

A multimodal anomaly de- tector for robot-assisted feeding using an lstm-based variational autoencoder,

D. Park, Y. Hoshi, and C. C. Kemp, “A multimodal anomaly de- tector for robot-assisted feeding using an lstm-based variational autoencoder,” IEEE Robotics and Automation Letters , vol. 3, no. 3, pp. 1544–1551, 2018

work page 2018
[26]

Robust anomaly detection for multivariate time series through stochas- tic recurrent neural network,

Y. Su, Y. Zhao, C. Niu, R. Liu, W. Sun, and D. Pei, “Robust anomaly detection for multivariate time series through stochas- tic recurrent neural network,” in Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , 2019, pp. 2828–2837

work page 2019
[27]

Multivariate time series anomaly detection and interpretation using hierarchical inter-metric and temporal embedding,

Z. Li, Y. Zhao, J. Han, Y. Su, R. Jiao, X. Wen, and D. Pei, “Multivariate time series anomaly detection and interpretation using hierarchical inter-metric and temporal embedding,” in Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining , 2021, pp. 3220–3230

work page 2021
[28]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

work page 2017
[29]

Breaking the time-frequency granularity discrepancy in time-series anomaly detection,

Y. Nam, S. Yoon, Y. Shin, M. Bae, H. Song, J.-G. Lee, and B. S. Lee, “Breaking the time-frequency granularity discrepancy in time-series anomaly detection,” in Proceedings of the ACM Web Conference 2024, 2024, pp. 4204–4215

work page 2024
[30]

Memto: Memory-guided transformer for multivariate time series anomaly detection,

J. Song, K. Kim, J. Oh, and S. Cho, “Memto: Memory-guided transformer for multivariate time series anomaly detection,” Advances in Neural Information Processing Systems , vol. 36, pp. 57 947–57 963, 2023. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12

work page 2023
[31]

Drift doesn’t matter: Dynamic decomposition with diffusion reconstruction for unstable multivariate time series anomaly detection,

C. Wang, Z. Zhuang, Q. Qi, J. Wang, X. Wang, H. Sun, and J. Liao, “Drift doesn’t matter: Dynamic decomposition with diffusion reconstruction for unstable multivariate time series anomaly detection,” Advances in neural information processing systems, vol. 36, pp. 10 758–10 774, 2023

work page 2023
[32]

An encode-then-decompose approach to unsu- pervised time series anomaly detection on contaminated train- ing data–extended version,

B. Zhang, T. Kieu, X. Qiu, C. Guo, J. Hu, A. Zhou, C. S. Jensen, and B. Yang, “An encode-then-decompose approach to unsu- pervised time series anomaly detection on contaminated train- ing data–extended version,” arXiv preprint arXiv:2510.18998 , 2025

work page arXiv 2025
[33]

NICE: Non-linear Independent Components Estimation

L. Dinh, D. Krueger, and Y. Bengio, “Nice: Non-linear indepen- dent components estimation,” arXiv preprint arXiv:1410.8516 , 2014

work page internal anchor Pith review Pith/arXiv arXiv 2014
[34]

Masked au- toregressive flow for density estimation,

G. Papamakarios, T. Pavlakou, and I. Murray, “Masked au- toregressive flow for density estimation,” Advances in neural information processing systems , vol. 30, 2017

work page 2017
[35]

Glow: Generative flow with invertible 1x1 convolutions,

D. P. Kingma and P. Dhariwal, “Glow: Generative flow with invertible 1x1 convolutions,” Advances in neural information processing systems, vol. 31, 2018

work page 2018
[36]

Neural ordinary differential equations,

R. T. Chen, Y. Rubanova, J. Bettencourt, and D. K. Duvenaud, “Neural ordinary differential equations,” Advances in neural information processing systems , vol. 31, 2018

work page 2018
[37]

Neu- ral spline flows,

C. Durkan, A. Bekasov, I. Murray, and G. Papamakarios, “Neu- ral spline flows,” Advances in neural information processing systems, vol. 32, 2019

work page 2019
[38]

Neural discrete repre- sentation learning,

A. Van Den Oord, O. Vinyals et al. , “Neural discrete repre- sentation learning,” Advances in neural information processing systems, vol. 30, 2017

work page 2017
[39]

Made: Masked autoencoder for distribution estimation,

M. Germain, K. Gregor, I. Murray, and H. Larochelle, “Made: Masked autoencoder for distribution estimation,” in Interna- tional conference on machine learning . PMLR, 2015, pp. 881– 889

work page 2015
[40]

Isolation forest,

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

work page 2008
[41]

Joint selective state space model and detrending for robust time series anomaly detection,

J. Chen, X. Tan, S. Rahardja, J. Yang, and S. Rahardja, “Joint selective state space model and detrending for robust time series anomaly detection,” IEEE Signal Processing Letters , 2024

work page 2024
[42]

Swat: A water treatment testbed for research and training on ics security,

A. P. Mathur and N. O. Tippenhauer, “Swat: A water treatment testbed for research and training on ics security,” in 2016 inter- national workshop on cyber-physical systems for smart water networks (CySWater). IEEE, 2016, pp. 31–36

work page 2016
[43]

Practical approach to asynchronous multivariate time series anomaly detection and localization,

A. Abdulaal, Z. Liu, and T. Lancewicki, “Practical approach to asynchronous multivariate time series anomaly detection and localization,” in Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining , 2021, pp. 2485–2494

work page 2021
[44]

Swan-sf: Space weather analytics dataset-solar flares,

R. Angryk, P. Martens, B. Aydin, D. Kempton, S. Mahajan, S. Basodi, A. Ahmadzadeh, X. Cai, S. Filali Boubrahimi, S. M. Hamdi et al. , “Swan-sf: Space weather analytics dataset-solar flares,” Harvard Dataverse dataset , p. 102, 2020

work page 2020
[45]

Towards a rigorous evaluation of time-series anomaly detection,

S. Kim, K. Choi, H.-S. Choi, B. Lee, and S. Yoon, “Towards a rigorous evaluation of time-series anomaly detection,” in Proceedings of the AAAI conference on artificial intelligence , vol. 36, no. 7, 2022, pp. 7194–7201

work page 2022
[46]

Local evaluation of time series anomaly detection algorithms,

A. Huet, J. M. Navarro, and D. Rossi, “Local evaluation of time series anomaly detection algorithms,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022, pp. 635–645

work page 2022
[47]

Density estimation using Real NVP

L. Dinh, J. Sohl-Dickstein, and S. Bengio, “Density estimation using real nvp,” arXiv preprint arXiv:1605.08803 , 2016

work page internal anchor Pith review Pith/arXiv arXiv 2016
[48]

Tsb-uad: an end-to-end benchmark suite for univariate time-series anomaly detection,

J. Paparrizos, Y. Kang, P. Boniol, R. S. Tsay, T. Palpanas, and M. J. Franklin, “Tsb-uad: an end-to-end benchmark suite for univariate time-series anomaly detection,” Proceedings of the VLDB Endowment , vol. 15, no. 8, pp. 1697–1711, 2022

work page 2022