Pith. sign in

REVIEW 2 major objections 2 minor 63 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

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

T0 review · grok-4.3

A hypergraph energy model learns the normal world from healthy data and calibrates boundaries with few abnormal examples to detect departures.

2026-06-26 11:43 UTC pith:PZ72XDI4

load-bearing objection The paper frames abnormality detection as learning a normal world via hypergraphs and a three-term entropy energy, with strong C-MAPSS numbers, but the full methods are needed to check calibration and independence of terms. the 2 major comments →

arxiv 2606.22261 v1 pith:PZ72XDI4 submitted 2026-06-20 cs.LG cs.GT

Learning a Normal World Model for Few-Shot Boundary-Calibrated Abnormality Detection

Weizhi Nie , Weichao Liu , Weijie Wang , Yuting Su This is my paper
classification cs.LG cs.GT
keywords abnormality detectionnormal world modelhypergraphfew-shot learningenergy-based modelanomaly detectionC-MAPSSturbofan degradation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper establishes that abnormality detection can proceed by first learning a complete representation of normal system behavior from abundant normal sensor data rather than attempting to catalog abnormal classes. It instantiates this as an entropy-aware energy function on context-conditioned hypergraphs that measures surprise in temporal predictions, relational consistency, and manifold position. Few abnormal examples are used only to set the decision threshold after the normal model is fixed. On the NASA C-MAPSS turbofan benchmark the resulting scores achieve strong zero-shot and few-shot performance, including an AUROC of 0.9983 on the most complex multi-condition subset. This matters because normal operation data is cheap to collect while labeled faults remain scarce, so a method that inverts the usual labeling burden offers a practical route to reliable detection.

Core claim

The central claim is that an entropy-aware normal-world energy defined on context-conditioned hypergraphs, formed by combining temporal prediction surprise, hypergraph consistency surprise, and latent normal-manifold departure, can serve simultaneously as an anomaly score, a graded risk measure, and a testable representation of normal system behavior, reaching an AUROC of 0.9983 on FD004 of the C-MAPSS dataset under few-shot boundary calibration.

What carries the argument

The Hypergraph Entropic Normal-World Model, which encodes multivariate sensor windows as context-conditioned hypergraphs and computes abnormality via an energy that aggregates multiple surprise terms to quantify departure from the learned normal world.

Load-bearing premise

Abundant normal events are sufficient to learn a representative normal world, and a small number of abnormal examples can set the decision boundary without introducing bias or overfitting.

What would settle it

A demonstration that the energy scores fail to increase monotonically along known degradation trajectories or fail to sharply penalize artificially introduced context-mismatched cross-variable couplings would show that the energy does not encode normal-world structure.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The energy functions as both an anomaly score and a graded risk measure that rises along degradation trajectories.
  • The learned model accepts unseen healthy engines while rejecting context-mismatched hypergraph configurations.
  • Mechanistic validation tests confirm that the energy captures normal-world structure rather than a superficial input-output mapping.
  • Normal-world energy can serve as a testable representation of normal system behavior under severe abnormal-label scarcity.

Where Pith is reading between the lines

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

  • If the hypergraph structure preserves identifiable variable groups, the same energy could support root-cause localization by highlighting which hyperedges contribute most to high scores.
  • The calibration step may be sensitive to which abnormal examples are chosen; repeating boundary calibration across multiple small abnormal subsets would test robustness.
  • The formulation could apply directly to other sensor-rich domains such as medical device monitoring where normal recordings vastly outnumber labeled faults.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper proposes a Hypergraph Entropic Normal-World Model for few-shot boundary-calibrated abnormality detection. It learns a normal world from abundant normal events by representing multivariate sensor windows as context-conditioned hypergraphs and defines an entropy-aware energy combining temporal prediction surprise, hypergraph consistency surprise, and latent normal-manifold departure. Few abnormal examples calibrate only the normality boundary. On NASA C-MAPSS, the full energy achieves strong zero- and few-shot AUROC across subsets, reaching 0.9983 on FD004, with mechanistic validation tests showing the energy accepts healthy engines, increases along degradation, and penalizes context-mismatched couplings.

Significance. If the energy formulation and calibration procedure hold without leakage or overfitting, the work offers a principled shift from modeling rare abnormalities to learning and testing a normal world, which could advance interpretable anomaly detection in safety-critical multivariate time series. The mechanistic validation tests are a clear strength, providing falsifiable probes beyond standard metrics, and the hypergraph representation for high-order relations is a substantive technical choice.

major comments (2)
  1. [Methods (energy definition and calibration procedure)] The entropy-aware energy (combining the three surprise terms) and its relation to the few-shot boundary calibration step require explicit equations and pseudocode; without them the claim that calibration uses abnormal examples only for boundary setting (and does not introduce circular dependence) cannot be verified, directly affecting the central few-shot claim.
  2. [Experiments (FD004 results and data protocol)] Table or figure reporting FD004 results (AUROC 0.9983): the data splits, operating-condition encoding, and handling of multiple fault modes must be shown to rule out leakage between the normal-world training set and the few abnormal calibration examples; the current description leaves open whether the high performance is robust or split-dependent.
minor comments (2)
  1. [Abstract] The abstract states 'mechanistic validation tests' but does not list the exact tests or quantitative thresholds; a short enumerated list would improve clarity.
  2. [Methods] Notation for the hypergraph construction (context-conditioned hyperedges) should be introduced with a small illustrative diagram or one-sentence definition on first use.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment of the mechanistic validation tests and for highlighting the potential significance of the normal-world modeling approach. Below we respond point-by-point to the two major comments. We agree that additional explicit detail is warranted on both the energy formulation and the data protocol, and we will incorporate the requested material in the revision.

read point-by-point responses

  1. Referee: [Methods (energy definition and calibration procedure)] The entropy-aware energy (combining the three surprise terms) and its relation to the few-shot boundary calibration step require explicit equations and pseudocode; without them the claim that calibration uses abnormal examples only for boundary setting (and does not introduce circular dependence) cannot be verified, directly affecting the central few-shot claim.

    Authors: We agree that the current manuscript description of the energy is insufficiently explicit for independent verification. In the revised version we will insert the full set of equations defining the three surprise terms (temporal prediction surprise, hypergraph consistency surprise, and latent normal-manifold departure) and their weighted sum into the energy function (new Equations 3–7 in Section 3.2). We will also add Algorithm 1 (pseudocode) that shows the two-stage procedure: (i) training the hypergraph normal-world model exclusively on normal data, followed by (ii) computing energy scores on the held-out few abnormal examples solely to select the decision threshold. No parameters of the normal-world model are updated during calibration, eliminating circular dependence. These additions directly address the verifiability concern. revision: yes

  2. Referee: [Experiments (FD004 results and data protocol)] Table or figure reporting FD004 results (AUROC 0.9983): the data splits, operating-condition encoding, and handling of multiple fault modes must be shown to rule out leakage between the normal-world training set and the few abnormal calibration examples; the current description leaves open whether the high performance is robust or split-dependent.

    Authors: We acknowledge that the current text does not provide a sufficiently granular account of the FD004 protocol. In the revision we will add a dedicated table (Table 2) and accompanying text in Section 4.1 that explicitly lists: (a) the exact train / calibration / test partition sizes and indices for each operating condition, (b) the encoding used for the six operating conditions (one-hot vectors concatenated to the sensor window), and (c) the procedure for sampling the few abnormal calibration examples from the multiple fault modes while ensuring zero overlap with the normal-world training engines. We will also report AUROC statistics over five independent random splits of the calibration set to demonstrate that the 0.9983 result is not an artifact of a single favorable partition. These changes will allow readers to confirm the absence of leakage. revision: yes

Circularity Check

0 steps flagged

No circularity detectable; insufficient equations to inspect derivation chain

full rationale

The provided abstract and context describe a Hypergraph Entropic Normal-World Model whose energy combines temporal prediction surprise, hypergraph consistency surprise, and latent normal-manifold departure, with few-shot boundary calibration on abnormal examples. No equations, self-citations, or fitted-parameter renamings appear in the text. Without the full manuscript's specific formulations (explicitly noted as unavailable), no load-bearing step can be shown to reduce by construction to its inputs. The approach is presented as validated on the external C-MAPSS benchmark with mechanistic tests, satisfying the default expectation of no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit list of free parameters, axioms, or invented entities; the hypergraph representation and energy function are introduced but their internal parameters and assumptions are not detailed.

pith-pipeline@v0.9.1-grok · 5800 in / 1216 out tokens · 24384 ms · 2026-06-26T11:43:40.952339+00:00 · methodology

0 comments
read the original abstract

Abnormality detection in complex systems faces two practical barriers: abnormal labels are scarce, and binary labels do not quantify how far an event has departed from normal behavior. We study a normal-world modeling formulation for this setting. Instead of learning a large and incomplete space of abnormal classes, the model learns the normal world from abundant normal events and uses a few abnormal examples only to calibrate the boundary of normality. We instantiate this idea as a Hypergraph Entropic Normal-World Model. The model represents multivariate sensor windows as context-conditioned hypergraphs, where hyperedges capture high-order relations among groups of variables. It then defines abnormality by an entropy-aware normal-world energy that combines temporal prediction surprise, hypergraph consistency surprise, and latent normal-manifold departure. On the NASA C-MAPSS turbofan degradation benchmark, the proposed full energy achieves strong zero-shot and few-shot performance across all four subsets and reaches AUROC 0.9983 on FD004, the most complex setting with multiple operating conditions and fault modes. Beyond standard detection metrics, we introduce mechanistic validation tests to probe whether the energy encodes normal-world structure rather than a superficial input-output mapping. The learned energy accepts unseen healthy engines, increases along degradation trajectories, and sharply penalizes context-mismatched cross-variable coupling breaks. These results suggest that normal-world energy can serve as an anomaly score, a graded risk measure, and a testable representation of normal system behavior under severe abnormal-label scarcity.

Figures

Figures reproduced from arXiv: 2606.22261 by Weichao Liu, Weijie Wang, Weizhi Nie, Yuting Su.

Figure 1
Figure 1. Figure 1: Motivation. Abnormal-class learning is label hungry, while a normal world model first [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Framework overview. A multivariate normal event window is represented as a context [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hypergraph construction. Sensors are represented as nodes, operating context modulates [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Few-shot boundary calibration compared with the strongest traditional and neural baselines [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Normal-world energy accepts held-out healthy engines and increases along degradation [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Cross-variable coupling break. Each row corresponds to one C-MAPSS subset; the [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Few-shot boundary calibration and context counterfactual consistency. The top row [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Hyperparameter sensitivity curves on FD002 and FD004. AUROC is shown for model [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗

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

63 extracted references · 8 canonical work pages · 5 internal anchors

  1. [1]

    Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture

    Assran, M., Duval, Q., Misra, I., Bojanowski, P., Vincent, P., Rabbat, M., LeCun, Y., and Ballas, N. Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023

    2023

  2. [2]

    Audibert, J., Michiardi, P., Guyard, F., Marti, S., and Zuluaga, M. A. USAD: UnSupervised Anomaly Detection on Multivariate Time Series. arXiv preprint arXiv:2006.10674, 2020

    work page arXiv 2006

  3. [3]

    An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling

    Bai, S., Kolter, J. Z., and Koltun, V. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling. arXiv preprint arXiv:1803.01271, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  4. [4]

    Revisiting Feature Prediction for Learning Visual Representations from Video

    Bardes, A., Garrido, Q., Ponce, J., Chen, X., Rabbat, M., LeCun, Y., Assran, M., and Ballas, N. Revisiting Feature Prediction for Learning Visual Representations from Video. arXiv preprint arXiv:2404.08471, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  5. [5]

    Spectral Temporal Graph Neural Network for Multivariate Time-Series Forecasting

    Cao, D., Wang, Y., Duan, J., Zhang, C., Zhu, X., Huang, C., Tong, Y., Xu, B., Bai, J., Tong, J., and Zhang, Q. Spectral Temporal Graph Neural Network for Multivariate Time-Series Forecasting. In Advances in Neural Information Processing Systems, 2020

    2020

  6. [6]

    Anomaly Detection: A Survey

    Chandola, V., Banerjee, A., and Kumar, V. Anomaly Detection: A Survey. ACM Computing Surveys, 41(3), 2009

    2009

  7. [7]

    Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation

    Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., and Bengio, Y. Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2014

    2014

  8. [8]

    You Are AllSet: A Multiset Function Framework for Hypergraph Neural Networks

    Chien, E., Pan, C., Peng, J., and Milenkovic, O. You Are AllSet: A Multiset Function Framework for Hypergraph Neural Networks. In International Conference on Learning Representations, 2022

    2022

  9. [9]

    and Hart, P

    Cover, T. and Hart, P. Nearest Neighbor Pattern Classification. IEEE Transactions on Information Theory, 13(1):21--27, 1967

    1967

  10. [10]

    and Hooi, B

    Deng, A. and Hooi, B. Graph Neural Network-Based Anomaly Detection in Multivariate Time Series. In Proceedings of the AAAI Conference on Artificial Intelligence, 2021

    2021

  11. [11]

    Hypergraph Neural Networks

    Feng, Y., You, H., Zhang, Z., Ji, R., and Gao, Y. Hypergraph Neural Networks. In Proceedings of the AAAI Conference on Artificial Intelligence, 2019

    2019

  12. [12]

    TadGAN: Time Series Anomaly Detection Using Generative Adversarial Networks

    Geiger, A., Liu, D., Alnegheimish, S., Cuesta-Infante, A., and Veeramachaneni, K. TadGAN: Time Series Anomaly Detection Using Generative Adversarial Networks. In Proceedings of the IEEE International Conference on Big Data, 2020

    2020

  13. [13]

    MOMENT: A Family of Open Time-Series Foundation Models

    Goswami, M., Szafer, K., Choudhry, A., Cai, Y., Li, S., and Dubrawski, A. MOMENT: A Family of Open Time-Series Foundation Models. In International Conference on Machine Learning, 2024

    2024

  14. [14]

    V., and Jain, P

    Goyal, S., Raghunathan, A., Jain, M., Simhadri, H. V., and Jain, P. DROCC: Deep Robust One-Class Classification. In International Conference on Machine Learning, 2020

    2020

  15. [15]

    Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One

    Grathwohl, W., Wang, K.-C., Jacobsen, J.-H., Duvenaud, D., Norouzi, M., and Swersky, K. Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One. In International Conference on Learning Representations, 2020

    2020

  16. [16]

    and Hyvarinen, A

    Gutmann, M. and Hyvarinen, A. Noise-Contrastive Estimation: A New Estimation Principle for Unnormalized Statistical Models. In Proceedings of the International Conference on Artificial Intelligence and Statistics, 2010

    2010

  17. [17]

    World Models

    Ha, D. and Schmidhuber, J. World Models. arXiv preprint arXiv:1803.10122, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  18. [18]

    Dream to Control: Learning Behaviors by Latent Imagination

    Hafner, D., Lillicrap, T., Ba, J., and Norouzi, M. Dream to Control: Learning Behaviors by Latent Imagination. In International Conference on Learning Representations, 2020

    2020

  19. [19]

    Mastering Diverse Domains through World Models

    Hafner, D., Pasukonis, J., Ba, J., and Lillicrap, T. Mastering Diverse Domains through World Models. arXiv preprint arXiv:2301.04104, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  20. [20]

    and Gimpel, K

    Hendrycks, D. and Gimpel, K. A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks. In International Conference on Learning Representations, 2017

    2017

  21. [21]

    Hoerl, A. E. and Kennard, R. W. Ridge Regression: Biased Estimation for Nonorthogonal Problems. Technometrics, 12(1):55--67, 1970

    1970

  22. [22]

    Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding

    Hundman, K., Constantinou, V., Laporte, C., Colwell, I., and Soderstrom, T. Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding. In Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2018

    2018

  23. [23]

    Estimation of Non-Normalized Statistical Models by Score Matching

    Hyvarinen, A. Estimation of Non-Normalized Statistical Models by Score Matching. Journal of Machine Learning Research, 6:695--709, 2005

    2005

  24. [24]

    Jolliffe, I. T. Principal Component Analysis. Springer, 2nd edition, 2002

    2002

  25. [25]

    Equivariant Hypergraph Neural Networks

    Kim, J., Oh, S., and Kim, S. Equivariant Hypergraph Neural Networks. In European Conference on Computer Vision, 2022

    2022

  26. [26]

    LeCun, Y., Chopra, S., Hadsell, R., Ranzato, M., and Huang, F. J. A Tutorial on Energy-Based Learning. In Predicting Structured Data. MIT Press, 2006

    2006

  27. [27]

    MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks

    Li, D., Chen, D., Shi, L., Jin, B., Goh, J., and Ng, S.-K. MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks. In International Conference on Artificial Neural Networks, 2019

    2019

  28. [28]

    T., Ting, K

    Liu, F. T., Ting, K. M., and Zhou, Z.-H. Isolation Forest. In Proceedings of the IEEE International Conference on Data Mining, 2008

    2008

  29. [29]

    Energy-Based Out-of-Distribution Detection

    Liu, W., Wang, X., Owens, J., and Li, Y. Energy-Based Out-of-Distribution Detection. In Advances in Neural Information Processing Systems, 2020

    2020

  30. [30]

    A., Franks, B

    Liznerski, P., Ruff, L., Vandermeulen, R. A., Franks, B. J., Kloft, M., and Muller, K.-R. Explainable Deep One-Class Classification. In International Conference on Learning Representations, 2021

    2021

  31. [31]

    Mahalanobis, P. C. On the Generalized Distance in Statistics. Proceedings of the National Institute of Sciences of India, 2(1):49--55, 1936

    1936

  32. [32]

    LSTM-based Encoder-Decoder for Multi-sensor Anomaly Detection

    Malhotra, P., Ramakrishnan, A., Anand, G., Vig, L., Agarwal, P., and Shroff, G. LSTM-Based Encoder-Decoder for Multi-Sensor Anomaly Detection. arXiv preprint arXiv:1607.00148, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  33. [33]

    Multi-grained Random Fields for Mitosis Identification in Time-Lapse Phase Contrast Microscopy Image Sequences

    Liu, A.-A., Tang, J., Nie, W., and Su, Y. Multi-grained Random Fields for Mitosis Identification in Time-Lapse Phase Contrast Microscopy Image Sequences. IEEE Transactions on Medical Imaging, 36(8):1699--1710, 2017

    2017

  34. [34]

    MF-GCN: Multimodal Information Fusion Using Incremental Graph Convolutional Network for Ship Behavior Anomaly Detection

    Ma, R., Zhang, J., Nie, W., Ge, N., Wen, H., and Liu, A. MF-GCN: Multimodal Information Fusion Using Incremental Graph Convolutional Network for Ship Behavior Anomaly Detection. Journal of Marine Science and Engineering, 14(1):87, 2026

    2026

  35. [35]

    Modeling Temporal Information of Mitotic for Mitotic Event Detection

    Nie, W., Cheng, H., and Su, Y. Modeling Temporal Information of Mitotic for Mitotic Event Detection. IEEE Transactions on Big Data, 3(4):458--469, 2017

    2017

  36. [36]

    Hyper-clique Graph Matching and Applications

    Nie, W.-Z., Liu, A.-A., Gao, Y., and Su, Y.-T. Hyper-clique Graph Matching and Applications. IEEE Transactions on Circuits and Systems for Video Technology, 29(6):1619--1630, 2018

    2018

  37. [37]

    Subgraph Learning for Graph Matching

    Nie, W., Ding, H., Liu, A., Deng, Z., and Su, Y. Subgraph Learning for Graph Matching. Pattern Recognition Letters, 130:362--369, 2020

    2020

  38. [38]

    Hierarchical Graph Structure Learning for Multi-View 3D Model Retrieval

    Su, Y., Li, W., Liu, A., and Nie, W. Hierarchical Graph Structure Learning for Multi-View 3D Model Retrieval. In Proceedings of the International Joint Conference on Artificial Intelligence, pages 913--919, 2018

    2018

  39. [39]

    A Multiscale Graph Convolutional Neural Network Framework for Fault Diagnosis of Rolling Bearing

    Yin, P., Nie, J., Liang, X., Yu, S., Wang, C., Nie, W., and Ding, X. A Multiscale Graph Convolutional Neural Network Framework for Fault Diagnosis of Rolling Bearing. IEEE Transactions on Instrumentation and Measurement, 72:1--13, 2023

    2023

  40. [40]

    H., Sinthong, P., and Kalagnanam, J

    Nie, Y., Nguyen, N. H., Sinthong, P., and Kalagnanam, J. A Time Series is Worth 64 Words: Long-Term Forecasting with Transformers. In International Conference on Learning Representations, 2023

    2023

  41. [41]

    Deep Anomaly Detection with Deviation Networks

    Pang, G., Shen, C., and van den Hengel, A. Deep Anomaly Detection with Deviation Networks. In Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2019

    2019

  42. [42]

    Deep Learning for Anomaly Detection: A Review

    Pang, G., Shen, C., Cao, L., and van den Hengel, A. Deep Learning for Anomaly Detection: A Review. ACM Computing Surveys, 54(2), 2021

    2021

  43. [43]

    GLUE: A Graph-Based Uncertainty Estimation Framework for Multivariate Time Series Anomaly Detection

    Ray, A., Sahu, S., Kumar, A., and Veeramachaneni, K. GLUE: A Graph-Based Uncertainty Estimation Framework for Multivariate Time Series Anomaly Detection. arXiv preprint arXiv:2106.02707, 2021

    work page arXiv 2021

  44. [44]

    A., Goernitz, N., Deecke, L., Siddiqui, S

    Ruff, L., Vandermeulen, R. A., Goernitz, N., Deecke, L., Siddiqui, S. A., Binder, A., Muller, E., and Kloft, M. Deep One-Class Classification. In International Conference on Machine Learning, 2018

    2018

  45. [45]

    A., Goernitz, N., Binder, A., Muller, E., Muller, K.-R., and Kloft, M

    Ruff, L., Vandermeulen, R. A., Goernitz, N., Binder, A., Muller, E., Muller, K.-R., and Kloft, M. Deep Semi-Supervised Anomaly Detection. In International Conference on Learning Representations, 2020

    2020

  46. [46]

    Damage Propagation Modeling for Aircraft Engine Run-to-Failure Simulation

    Saxena, A., Goebel, K., Simon, D., and Eklund, N. Damage Propagation Modeling for Aircraft Engine Run-to-Failure Simulation. NASA Ames Research Center, 2008

    2008

  47. [47]

    C., Shawe-Taylor, J., Smola, A

    Scholkopf, B., Platt, J. C., Shawe-Taylor, J., Smola, A. J., and Williamson, R. C. Estimating the Support of a High-Dimensional Distribution. Neural Computation, 13(7):1443--1471, 2001

    2001

  48. [48]

    Shannon, C. E. A Mathematical Theory of Communication. Bell System Technical Journal, 27(3):379--423 and 27(4):623--656, 1948

    1948

  49. [49]

    Robust Anomaly Detection for Multivariate Time Series through Stochastic Recurrent Neural Network

    Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., and Pei, D. Robust Anomaly Detection for Multivariate Time Series through Stochastic Recurrent Neural Network. In Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2019

    2019

  50. [50]

    Tax, D. M. J. and Duin, R. P. W. Support Vector Data Description. Machine Learning, 54:45--66, 2004

    2004

  51. [51]

    Tuli, S., Casale, G., and Jennings, N. R. TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data. Proceedings of the VLDB Endowment, 15(6):1201--1214, 2022

    2022

  52. [52]

    N., Kaiser, L., and Polosukhin, I

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., and Polosukhin, I. Attention Is All You Need. In Advances in Neural Information Processing Systems, 2017

    2017

  53. [53]

    Graph WaveNet for Deep Spatial-Temporal Graph Modeling

    Wu, Z., Pan, S., Long, G., Jiang, J., and Zhang, C. Graph WaveNet for Deep Spatial-Temporal Graph Modeling. In Proceedings of the International Joint Conference on Artificial Intelligence, 2019

    2019

  54. [54]

    Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks

    Wu, Z., Pan, S., Long, G., Jiang, J., Chang, X., and Zhang, C. Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks. In Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2020

    2020

  55. [55]

    TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis

    Wu, H., Hu, T., Liu, Y., Zhou, H., Wang, J., and Long, M. TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis. In International Conference on Learning Representations, 2023

    2023

  56. [56]

    Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy

    Xu, J., Wu, H., Wang, J., and Long, M. Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy. In International Conference on Learning Representations, 2022

    2022

  57. [57]

    DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection

    Yang, Y., Zhang, C., Zhou, T., Wen, Q., and Sun, L. DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection. In Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023

    2023

  58. [58]

    DTAAD: Dual TCN-Attention Networks for Anomaly Detection in Multivariate Time Series Data

    Yu, L. DTAAD: Dual TCN-Attention Networks for Anomaly Detection in Multivariate Time Series Data. arXiv preprint arXiv:2302.10753, 2023

    work page arXiv 2023

  59. [59]

    Zhang, C., Song, D., Chen, Y., Feng, X., Lumezanu, C., Cheng, W., Ni, J., Zong, B., Chen, H., and Chawla, N. V. A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data. In Proceedings of the AAAI Conference on Artificial Intelligence, 2019

    2019

  60. [60]

    Hyper-SAGNN: A Self-Attention Based Graph Neural Network for Hypergraphs

    Zhang, R., Zou, Y., and Ma, J. Hyper-SAGNN: A Self-Attention Based Graph Neural Network for Hypergraphs. In International Conference on Learning Representations, 2020

    2020

  61. [61]

    Multivariate Time-Series Anomaly Detection via Graph Attention Network

    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. In Proceedings of the IEEE International Conference on Data Mining, 2020

    2020

  62. [62]

    Learning with Hypergraphs: Clustering, Classification, and Embedding

    Zhou, D., Huang, J., and Scholkopf, B. Learning with Hypergraphs: Clustering, Classification, and Embedding. In Advances in Neural Information Processing Systems, 2006

    2006

  63. [63]

    R., Cheng, W., Lumezanu, C., Cho, D., and Chen, H

    Zong, B., Song, Q., Min, M. R., Cheng, W., Lumezanu, C., Cho, D., and Chen, H. Deep Autoencoding Gaussian Mixture Model for Unsupervised Anomaly Detection. In International Conference on Learning Representations, 2018

    2018