pith. sign in

arxiv: 2604.14726 · v2 · submitted 2026-04-16 · 💻 cs.LG · cs.AI

Catching Every Ripple: Enhanced Anomaly Awareness via Dynamic Concept Adaptation

Jiaqi Zhu , Shaofeng Cai , Jie Chen , Fang Deng , Beng Chin Ooi , Wenqiao Zhang This is my paper

Pith reviewed 2026-05-10 12:28 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords online anomaly detectionconcept driftdynamic adaptationhypernetworkparameter shiftingdynamic thresholdingevolving data streamsanomaly awareness
0
0 comments X

The pith

DyMETER adapts online anomaly detectors to concept drift by generating instance-specific parameter shifts and recalibrating decision thresholds dynamically.

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

The paper presents DyMETER to improve online anomaly detection by allowing detectors to adapt to changing concepts in data streams. It trains an initial static model on past data and then applies a hypernetwork to shift parameters for new instances while using an evolution controller to manage uncertainty in those shifts. A dynamic threshold component further adjusts the boundary using samples whose status is unclear, enabling continuous alignment without costly retraining. This integrated online approach is intended to deliver stronger performance than existing methods that either stay rigid or require repeated full updates.

Core claim

DyMETER first trains a static detector on historical data and then employs a hypernetwork to produce instance-aware parameter shifts for adaptation to new concepts, combined with an evolution controller that estimates instance-level uncertainty to guide updates and a dynamic threshold module that uses a window of uncertain samples to adjust boundaries, all within one online process for handling concept drift in anomaly detection.

What carries the argument

The dynamic concept adaptation mechanism that uses a hypernetwork for instance-aware parameter shifts, an evolution controller for uncertainty estimation, and a candidate window for threshold recalibration.

If this is right

  • Detectors respond to emerging concepts during operation without full retraining.
  • Decision boundaries stay aligned with current data patterns through ongoing recalibration.
  • Uncertainty estimates make the adaptation steps more interpretable and controlled.
  • The unified online process covers a wider range of drifting stream scenarios than rigid or retraining-based alternatives.

Where Pith is reading between the lines

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

  • Separating a stable base detector from on-the-fly adjustments could transfer to other streaming tasks such as classification or forecasting.
  • The uncertainty-driven updates might combine naturally with selective querying of borderline samples to refine the model further.
  • Long-running monitoring systems could operate with reduced retraining overhead if the adaptation remains stable across extended periods.

Load-bearing premise

The hypernetwork can reliably generate instance-aware parameter shifts for the static detector and the evolution controller can accurately estimate instance-level concept uncertainty to drive effective adaptation without introducing instability.

What would settle it

Testing DyMETER on a data stream that contains abrupt, unpredictable concept changes and checking whether detection accuracy stays higher than baselines or drops because of incorrect parameter shifts or miscalibrated thresholds.

Figures

Figures reproduced from arXiv: 2604.14726 by Beng Chin Ooi, Fang Deng, Jiaqi Zhu, Jie Chen, Shaofeng Cai, Wenqiao Zhang.

Figure 1
Figure 1. Figure 1: Adaptation paradigms of DyMETER and existing approaches for handling concept drift in evolving data streams. generating instance-aware parameter shifts via a hypernet￾work, avoiding continuous retraining or finetuning. • Extensive experiments demonstrate that DyMETER con￾sistently outperforms state-of-the-art OAD baselines across diverse drift scenarios. The remainder of this paper is organized as follows:… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed DyMETER. (a) Static Concept-aware Detector (SCD) is trained on historical data to model the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Ablation analysis on module effectiveness via AU￾CROC and AUCPR for DyMETER and its six variants. TABLE VIII: Performance on dedicated ablation studies mea￾sured by AUCROC. Variant Ion. BGL M.T. INSECTS INSECTS INSECTS -Mix -Abr -Inc DyMETER-lstm 0.976 0.925 0.912 0.815 0.918 0.859 DyMETER-conv 0.982 0.924 0.893 0.811 0.889 0.838 DyMETER-r-conv 0.986 0.932 0.892 0.807 0.916 0.841 DyMETER-ce 0.961 0.898 0.8… view at source ↗
Figure 4
Figure 4. Figure 4: Sensitivity test of key hyperparameters. Concept A Concept B Concept C Concept D Concept E t=16000 Concept B t=16445 Concept Drift t=20000 Concept C Evolution Evolution Evolution Final Result: Normal Final Result: Normal Final Result: Normal 𝑃" = [0.8896, 0.1104], 𝒰!"# = 0.0086 𝑃" = [0.4761, 0.5239], 𝒰!"# = 0.0268 𝑃" = [0.8274, 0.1726], 𝒰!"# = 0.0092 Normal Abnormal Normal Abnormal Normal Abnormal [PITH_F… view at source ↗
Figure 5
Figure 5. Figure 5: Interpretation of model behavior across evolving concepts. within the sliding window WN , shows mild fluctuations but remains generally stable. Smaller µo values improve detection accuracy but may trigger frequent offline updates, thereby reducing efficiency. In practice, when prior data characteristics are unknown, we recommend setting µp between 0.1 and 0.2, µe within 0.01–0.05, and µo between 0.2 and 0.… view at source ↗
Figure 6
Figure 6. Figure 6: Robustness under noisy and insufficient concept coverage. concept uncertainty indicates a distributional shift, activating the IEC’s dynamic mode. At t = 16445, DyMETER yields a probability distribution marked by significant concept uncer￾tainty, suggesting a potential for concept drift that warrants attention. In response, DyMETER transitions to the dynamic mode and adjusts base parameters in real time, a… view at source ↗
Figure 7
Figure 7. Figure 7: In-depth analysis on concept drift adaptation. to evolving concepts. To further assess adaptability under more complex conditions, we construct a large-scale dataset with mixed drift types by sequentially concatenating four INSECTS variants. This setup produces a unified data stream where multiple drift types (i.e., abrupt, incremental, gradual, and reoccurring) co-exist. To ensure label-space consistency … view at source ↗
Figure 8
Figure 8. Figure 8: Timely performance of DyMETER on Flink. [58]. With the rise of deep learning, reconstruction-based approaches leveraging autoencoders [59], [52], variational autoencoders (VAEs) [60], [56], and generative adversarial networks (GANs) [61] have become dominant, detecting anomalies as instances with large reconstruction errors. More recently, diffusion-based models [62], [63], [64] have emerged, further advan… view at source ↗
read the original abstract

Online anomaly detection (OAD) plays a pivotal role in real-time analytics and decision-making for evolving data streams. However, existing methods often rely on costly retraining and rigid decision boundaries, limiting their ability to adapt both effectively and efficiently to concept drift in dynamic environments. To address these challenges, we propose DyMETER, a dynamic concept adaptation framework for OAD that unifies on-the-fly parameter shifting and dynamic thresholding within a single online paradigm. DyMETER first learns a static detector on historical data to capture recurring central concepts, and then transitions to a dynamic mode to adapt to new concepts as drift occurs. Specifically, DyMETER employs a novel dynamic concept adaptation mechanism that leverages a hypernetwork to generate instance-aware parameter shifts for the static detector, thereby enabling efficient and effective adaptation without retraining or fine-tuning. To achieve robust and interpretable adaptation, DyMETER introduces a lightweight evolution controller to estimate instance-level concept uncertainty for adaptive updates. Further, DyMETER employs a dynamic threshold optimization module to adaptively recalibrates the decision boundary by maintaining a candidate window of uncertain samples, which ensures continuous alignment with evolving concepts. Extensive experiments demonstrate that DyMETER significantly outperforms existing OAD approaches across a wide spectrum of application scenarios.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 0 minor

Summary. The paper proposes DyMETER, a dynamic concept adaptation framework for online anomaly detection (OAD) in evolving data streams. It learns a static detector on historical data to capture central concepts, then employs a hypernetwork to generate instance-aware parameter shifts for the detector (enabling adaptation without retraining), a lightweight evolution controller to estimate instance-level concept uncertainty for driving updates, and a dynamic threshold optimization module that maintains a candidate window of uncertain samples to recalibrate decision boundaries. Extensive experiments are claimed to show significant outperformance over existing OAD methods across application scenarios.

Significance. If the mechanisms prove stable and the empirical gains hold under rigorous validation, DyMETER could offer a practical advance for real-time OAD by unifying parameter adaptation and thresholding in an online setting, reducing reliance on costly retraining while handling concept drift more responsively than rigid-boundary baselines.

major comments (2)
  1. Abstract: The central claim that the hypernetwork produces 'efficient and effective adaptation without retraining or fine-tuning' and that the evolution controller enables 'robust' updates rests on unstated assumptions about stability. No mechanism (e.g., regularization, bounds on parameter deltas, or Lipschitz constraints on the hypernetwork mapping) is described to prevent large or inconsistent shifts when input perturbations arise from drift, which directly risks violating the premise of reliable on-the-fly adaptation.
  2. Abstract: The dynamic threshold optimization module is said to 'adaptively recalibrates the decision boundary' via a 'candidate window of uncertain samples,' yet no details are given on window maintenance, update rules, or how it interacts with the uncertainty estimates from the evolution controller. This leaves the 'continuous alignment with evolving concepts' claim without a verifiable procedure.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and the recommendation for major revision. We address each major comment point by point below and will update the manuscript accordingly to improve clarity on stability and implementation details.

read point-by-point responses

  1. Referee: Abstract: The central claim that the hypernetwork produces 'efficient and effective adaptation without retraining or fine-tuning' and that the evolution controller enables 'robust' updates rests on unstated assumptions about stability. No mechanism (e.g., regularization, bounds on parameter deltas, or Lipschitz constraints on the hypernetwork mapping) is described to prevent large or inconsistent shifts when input perturbations arise from drift, which directly risks violating the premise of reliable on-the-fly adaptation.

    Authors: We agree that the current manuscript does not sufficiently describe explicit mechanisms to ensure stability of the hypernetwork-generated shifts. The abstract and methods sections summarize the adaptation process but omit details on regularization or bounds. In the revised manuscript we will add a dedicated paragraph in Section 3.2 describing a regularization term on the hypernetwork outputs to constrain parameter deltas, together with an empirical analysis of shift magnitudes across drift scenarios. We will also discuss the role of the evolution controller in gating updates to limit inconsistency. revision: yes

  2. Referee: Abstract: The dynamic threshold optimization module is said to 'adaptively recalibrates the decision boundary' via a 'candidate window of uncertain samples,' yet no details are given on window maintenance, update rules, or how it interacts with the uncertainty estimates from the evolution controller. This leaves the 'continuous alignment with evolving concepts' claim without a verifiable procedure.

    Authors: We agree that the description of the dynamic threshold optimization module lacks the necessary procedural specifics for verification. The manuscript provides only a high-level overview of the candidate window. In the revision we will expand Section 3.3 with explicit update rules (including window size, insertion criteria based on uncertainty scores, and FIFO maintenance), the precise interaction with the evolution controller, and the optimization objective used for boundary recalibration. Pseudocode for the full module will also be added. revision: yes

Circularity Check

0 steps flagged

No significant circularity; DyMETER components are independently constructed

full rationale

The paper describes DyMETER as a composite framework: a static detector pretrained on historical data, followed by a hypernetwork that generates instance-aware parameter shifts, a lightweight evolution controller estimating instance-level concept uncertainty, and a dynamic threshold module maintaining a candidate window of uncertain samples. None of these steps are shown to reduce by construction to the inputs via equations, self-definitions, or load-bearing self-citations. The abstract and framework overview present the elements as additive novel mechanisms for on-the-fly adaptation without retraining, with no renaming of known results or fitted quantities relabeled as predictions. The derivation chain remains self-contained against external benchmarks, consistent with the reader's assessment of no detectable circularity at the framework level.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no mathematical derivations, fitted parameters, background axioms, or new postulated entities are specified in the provided text.

pith-pipeline@v0.9.0 · 5533 in / 1079 out tokens · 68933 ms · 2026-05-10T12:28:41.855968+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages · 1 internal anchor

  1. [1]

    Self-supervised anomaly detection with neural transformations,

    C. Qiu, M. Kloft, S. Mandt, and M. Rudolph, “Self-supervised anomaly detection with neural transformations,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    work page 2024

  2. [2]

    A robust prioritized anomaly detection when not all anomalies are of primary interest,

    G. Lu, F. Zhou, M. Pavlovski, C. Zhou, and C. Jin, “A robust prioritized anomaly detection when not all anomalies are of primary interest,” in 2024 IEEE 40th International Conference on Data Engineering (ICDE). IEEE, 2024, pp. 775–788

    work page 2024

  3. [3]

    In- context adaptation to concept drift for learned database operations,

    J. Zhu, S. Cai, Y . Shen, G. Chen, F. Deng, and B. C. Ooi, “In- context adaptation to concept drift for learned database operations,” in International Conference on Machine Learning. PMLR, 2025, pp. 79 699–79 726

    work page 2025

  4. [4]

    METER: A dynamic concept adaptation framework for online anomaly detection,

    J. Zhu, S. Cai, F. Deng, B. C. Ooi, and W. Zhang, “METER: A dynamic concept adaptation framework for online anomaly detection,” Proceedings of the VLDB Endowment, vol. 17, no. 4, pp. 794–807, 2023

    work page 2023

  5. [5]

    Exploiting structural consistency of chest anatomy for unsupervised anomaly detection in radiography images,

    T. Xiang, Y . Zhang, Y . Lu, A. Yuille, C. Zhang, W. Cai, and Z. Zhou, “Exploiting structural consistency of chest anatomy for unsupervised anomaly detection in radiography images,”IEEE Transactions on Pat- tern Analysis and Machine Intelligence, 2024

    work page 2024

  6. [6]

    Rapp: Novelty detection with reconstruction along projection pathway,

    K. H. Kim, S. Shim, Y . Lim, J. Jeon, J. Choi, B. Kim, and A. S. Yoon, “Rapp: Novelty detection with reconstruction along projection pathway,” inInternational Conference on Learning Representations, 2020

    work page 2020

  7. [7]

    Robust subspace recovery layer for unsupervised anomaly detection,

    C.-H. Lai, D. Zou, and G. Lerman, “Robust subspace recovery layer for unsupervised anomaly detection,”arXiv preprint arXiv:1904.00152, 2019

    work page arXiv 1904

  8. [8]

    Adaptive model pooling for online deep anomaly detection from a complex evolving data stream,

    S. Yoon, Y . Lee, J.-G. Lee, and B. S. Lee, “Adaptive model pooling for online deep anomaly detection from a complex evolving data stream,” inProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022, pp. 2347–2357

    work page 2022

  9. [9]

    Kitsune: An ensemble of autoencoders for online network intrusion detection,

    Y . Mirsky, T. Doitshman, Y . Elovici, and A. Shabtai, “Kitsune: An ensemble of autoencoders for online network intrusion detection,” in 25th Annual Network and Distributed System Security Symposium, NDSS

    work page

  10. [10]

    The Internet Society, 2018

    work page 2018

  11. [11]

    Gaussian mixture variational autoencoder with whitening score for multimodal time series anomaly detection,

    J. Zhu, F. Deng, J. Zhao, Z. Ye, and J. Chen, “Gaussian mixture variational autoencoder with whitening score for multimodal time series anomaly detection,” in2022 IEEE 17th international conference on control & automation (ICCA). IEEE, 2022, pp. 480–485

    work page 2022

  12. [12]

    Mem- stream: Memory-based streaming anomaly detection,

    S. Bhatia, A. Jain, S. Srivastava, K. Kawaguchi, and B. Hooi, “Mem- stream: Memory-based streaming anomaly detection,” inProceedings of the ACM Web Conference 2022, 2022, pp. 610–621

    work page 2022

  13. [13]

    Mstream: Fast anomaly detection in multi-aspect streams,

    S. Bhatia, A. Jain, P. Li, R. Kumar, and B. Hooi, “Mstream: Fast anomaly detection in multi-aspect streams,” inProceedings of the Web Conference 2021, 2021, pp. 3371–3382

    work page 2021

  14. [14]

    Multiple dynamic outlier- detection from a data stream by exploiting duality of data and queries,

    S. Yoon, Y . Shin, J.-G. Lee, and B. S. Lee, “Multiple dynamic outlier- detection from a data stream by exploiting duality of data and queries,” inProceedings of the 2021 International Conference on Management of Data, 2021, pp. 2063–2075

    work page 2021

  15. [15]

    Robust random cut forest based anomaly detection on streams,

    S. Guha, N. Mishra, G. Roy, and O. Schrijvers, “Robust random cut forest based anomaly detection on streams,” inInternational conference on machine learning. PMLR, 2016, pp. 2712–2721

    work page 2016

  16. [16]

    Sand: streaming subsequence anomaly detection,

    P. Boniol, J. Paparrizos, T. Palpanas, and M. J. Franklin, “Sand: streaming subsequence anomaly detection,”Proceedings of the VLDB Endowment, vol. 14, no. 10, pp. 1717–1729, 2021

    work page 2021

  17. [17]

    Pidforest: anomaly detection via partial identification,

    P. Gopalan, V . Sharan, and U. Wieder, “Pidforest: anomaly detection via partial identification,”Advances in Neural Information Processing Systems, vol. 32, 2019

    work page 2019

  18. [18]

    Outlier detection for time series with recurrent autoencoder ensembles

    T. Kieu, B. Yang, C. Guo, and C. S. Jensen, “Outlier detection for time series with recurrent autoencoder ensembles.” inIJCAI, 2019, pp. 2725– 2732

    work page 2019

  19. [19]

    Drift doesn’t matter: Dynamic decomposition with diffusion reconstruc- tion 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 reconstruc- tion for unstable multivariate time series anomaly detection,”Advances in neural information processing systems, vol. 36, pp. 10 758–10 774, 2023

    work page 2023

  20. [20]

    Sarad: Spatial association-aware anomaly detection and diagnosis for multivariate time series,

    Z. Dai, L. He, S. Yang, and M. Leeke, “Sarad: Spatial association-aware anomaly detection and diagnosis for multivariate time series,”Advances in Neural Information Processing Systems, vol. 37, pp. 48 371–48 410, 2024

    work page 2024

  21. [21]

    State-transition-aware anomaly detec- tion under concept drifts,

    B. Li, S. Gupta, and E. M ¨uller, “State-transition-aware anomaly detec- tion under concept drifts,”Data & Knowledge Engineering, vol. 154, p. 102365, 2024

    work page 2024

  22. [22]

    When model meets new normals: Test-time adaptation for unsupervised time-series anomaly detection,

    D. Kim, S. Park, and J. Choo, “When model meets new normals: Test-time adaptation for unsupervised time-series anomaly detection,” in Proceedings of the AAAI conference on artificial intelligence, vol. 38, no. 12, 2024, pp. 13 113–13 121

    work page 2024

  23. [23]

    Evidential deep learning to quantify classification uncertainty,

    M. Sensoy, L. Kaplan, and M. Kandemir, “Evidential deep learning to quantify classification uncertainty,”Advances in neural information processing systems, vol. 31, 2018

    work page 2018

  24. [24]

    Rethinking unsupervised graph anomaly detection with deep learning: Residuals and objectives,

    X. Ma, F. Liu, J. Wu, J. Yang, S. Xue, and Q. Z. Sheng, “Rethinking unsupervised graph anomaly detection with deep learning: Residuals and objectives,”IEEE Transactions on Knowledge and Data Engineering, 2024

    work page 2024

  25. [25]

    Practical approach to asyn- chronous multivariate time series anomaly detection and localization,

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

    work page 2021

  26. [26]

    Denoising diffusion models for out-of-distribution detection,

    M. S. Graham, W. H. Pinaya, P.-D. Tudosiu, P. Nachev, S. Ourselin, and J. Cardoso, “Denoising diffusion models for out-of-distribution detection,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 2948–2957

    work page 2023

  27. [27]

    Statistical testing for efficient out of distribution detection in deep neural networks,

    M. Haroush, T. Frostig, R. Heller, and D. Soudry, “Statistical testing for efficient out of distribution detection in deep neural networks,”arXiv preprint arXiv:2102.12967, 2021

    work page arXiv 2021

  28. [28]

    Dirichlet and related distribu- tions: Theory, methods and applications,

    K. W. Ng, G.-L. Tian, and M.-L. Tang, “Dirichlet and related distribu- tions: Theory, methods and applications,” 2011

    work page 2011

  29. [29]

    Exponential moving average versus moving exponential average,

    F. Klinker, “Exponential moving average versus moving exponential average,”Mathematische Semesterberichte, vol. 58, pp. 97–107, 2011

    work page 2011

  30. [30]

    A meta-level analysis of online anomaly detectors,

    A. Ntroumpogiannis, M. Giannoulis, N. Myrtakis, V . Christophides, E. Simon, and I. Tsamardinos, “A meta-level analysis of online anomaly detectors,”The VLDB Journal, vol. 32, no. 4, pp. 845–886, 2023

    work page 2023

  31. [31]

    Distribution ag- nostic symbolic representations for time series dimensionality reduction and online anomaly detection,

    K. Bountrogiannis, G. Tzagkarakis, and P. Tsakalides, “Distribution ag- nostic symbolic representations for time series dimensionality reduction and online anomaly detection,”IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 6, pp. 5752–5766, 2022

    work page 2022

  32. [32]

    Jsang,Subjective Logic: A formalism for reasoning under uncertainty

    A. Jsang,Subjective Logic: A formalism for reasoning under uncertainty. Springer Publishing Company, Incorporated, 2018

    work page 2018

  33. [33]

    Odds library,

    S. Rayana, “Odds library,” 2016, https://odds.cs.stonybrook.edu. Accessed:2023-07

    work page 2016

  34. [34]

    What supercomputers say: A study of five system logs,

    A. Oliner and J. Stearley, “What supercomputers say: A study of five system logs,” in37th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN’07), 2007, pp. 575–584

    work page 2007

  35. [35]

    Kdd cup dataset,

    “Kdd cup dataset,” http://kdd.ics.uci.edu/databases/kddcup99/kddcup99. html. Accessed:2023-07, 1999

    work page 2023

  36. [36]

    A detailed analysis of the kdd cup 99 data set,

    M. Tavallaee, E. Bagheri, W. Lu, and A. A. Ghorbani, “A detailed analysis of the kdd cup 99 data set,” in2009 IEEE symposium on computational intelligence for security and defense applications. Ieee, 2009, pp. 1–6

    work page 2009

  37. [37]

    The ucr time series classification archive,

    H. A. Dau, E. Keogh, K. Kamgar, C.-C. M. Yeh, Y . Zhu, S. Gharghabi, C. A. Ratanamahatana, Yanping, B. Hu, N. Begum, A. Bagnall, A. Mueen, and G. Batista, “The ucr time series classification archive,” October 2021, https://www.cs.ucr.edu/∼eamonn/time series data 2018/ UCR TimeSeriesAnomalyDatasets2021.zip. Accessed:2023-07

    work page 2021

  38. [38]

    Current time series anomaly detection benchmarks are flawed and are creating the illusion of progress,

    R. Wu and E. J. Keogh, “Current time series anomaly detection benchmarks are flawed and are creating the illusion of progress,”IEEE transactions on knowledge and data engineering, vol. 35, no. 3, pp. 2421–2429, 2021

    work page 2021

  39. [39]

    Unsupervised real-time anomaly detection for streaming data,

    S. Ahmad, A. Lavin, S. Purdy, and Z. Agha, “Unsupervised real-time anomaly detection for streaming data,”Neurocomputing, vol. 262, pp. 134–147, 2017

    work page 2017

  40. [40]

    Challenges in benchmarking stream learning algorithms with real-world data,

    V . M. Souza, D. M. dos Reis, A. G. Maletzke, and G. E. Batista, “Challenges in benchmarking stream learning algorithms with real-world data,”Data Mining and Knowledge Discovery, vol. 34, pp. 1805–1858, 2020

    work page 2020

  41. [41]

    Gradient-based learning applied to document recognition,

    Y . LeCun, L. Bottou, Y . Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,”Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, 2002

    work page 2002

  42. [42]

    Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms

    H. Xiao, K. Rasul, and R. V ollgraf, “Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms,”arXiv preprint arXiv:1708.07747, 2017

    work page internal anchor Pith review arXiv 2017

  43. [43]

    Lof: identifying density-based local outliers,

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

    work page 2000

  44. [44]

    Isolation forest,

    F. T. Liu, K. M. Ting, and Z.-H. Zhou, “Isolation forest,” in2008 eighth ieee international conference on data mining. IEEE, 2008, pp. 413–422

    work page 2008

  45. [45]

    Detecting distance-based outliers in streams of data,

    F. Angiulli and F. Fassetti, “Detecting distance-based outliers in streams of data,” inProceedings of the sixteenth ACM conference on Conference on information and knowledge management, 2007, pp. 811–820

    work page 2007

  46. [46]

    Fast anomaly detection for streaming data,

    S. C. Tan, K. M. Ting, and T. F. Liu, “Fast anomaly detection for streaming data,” inTwenty-second international joint conference on artificial intelligence. Citeseer, 2011. THIS PAPER HAS BEEN ACCEPTED FOR PUBLICATION IN IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE (TPAMI). 16

    work page 2011

  47. [47]

    An anomaly detection approach based on isola- tion forest algorithm for streaming data using sliding window,

    Z. Ding and M. Fei, “An anomaly detection approach based on isola- tion forest algorithm for streaming data using sliding window,”IFAC Proceedings Volumes, vol. 46, no. 20, pp. 12–17, 2013

    work page 2013

  48. [48]

    Subspace outlier detection in linear time with randomized hashing,

    S. Sathe and C. C. Aggarwal, “Subspace outlier detection in linear time with randomized hashing,” in2016 IEEE 16th International Conference on Data Mining (ICDM). IEEE, 2016, pp. 459–468

    work page 2016

  49. [49]

    Loda: Lightweight on-line detector of anomalies,

    T. Pevn `y, “Loda: Lightweight on-line detector of anomalies,”Machine Learning, vol. 102, pp. 275–304, 2016

    work page 2016

  50. [50]

    xstream: Outlier detection in feature-evolving data streams,

    E. Manzoor, H. Lamba, and L. Akoglu, “xstream: Outlier detection in feature-evolving data streams,” inProceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2018, pp. 1963–1972

    work page 2018

  51. [51]

    Apache flink: Stream and batch processing in a single engine,

    P. Carbone, A. Katsifodimos, S. Ewen, V . Markl, S. Haridi, and K. Tzoumas, “Apache flink: Stream and batch processing in a single engine,”The Bulletin of the Technical Committee on Data Engineering, vol. 38, no. 4, 2015

    work page 2015

  52. [52]

    Intrusion detection scheme with dimensionality reduction in next generation networks,

    K. Sood, M. R. Nosouhi, D. D. N. Nguyen, F. Jiang, M. Chowdhury, and R. Doss, “Intrusion detection scheme with dimensionality reduction in next generation networks,”IEEE Transactions on Information Forensics and Security, vol. 18, pp. 965–979, 2023

    work page 2023

  53. [53]

    Adaptive aggregation-distillation autoencoder for unsupervised anomaly detection,

    J. Zhu, F. Deng, J. Zhao, and J. Chen, “Adaptive aggregation-distillation autoencoder for unsupervised anomaly detection,”Pattern Recognition, vol. 131, p. 108897, 2022

    work page 2022

  54. [54]

    Do llms understand visual anomalies? uncovering llm’s capabilities in zero-shot anomaly detection,

    J. Zhu, S. Cai, F. Deng, B. C. Ooi, and J. Wu, “Do llms understand visual anomalies? uncovering llm’s capabilities in zero-shot anomaly detection,” inProceedings of the 32nd ACM International Conference on Multimedia, 2024, pp. 48–57

    work page 2024

  55. [55]

    Few-shot domain-adaptive anomaly detection for cross-site brain images,

    J. Su, H. Shen, L. Peng, and D. Hu, “Few-shot domain-adaptive anomaly detection for cross-site brain images,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 46, no. 3, pp. 1819–1835, 2021

    work page 2021

  56. [56]

    Healthcare and anomaly detection: using machine learning to predict anomalies in heart rate data,

    E. ˇSabi´c, D. Keeley, B. Henderson, and S. Nannemann, “Healthcare and anomaly detection: using machine learning to predict anomalies in heart rate data,”AI & SOCIETY, vol. 36, no. 1, pp. 149–158, 2021

    work page 2021

  57. [57]

    UAED: Unsupervised abnormal emotion detection network based on wearable mobile device,

    J. Zhu, F. Deng, J. Zhao, D. Liu, and J. Chen, “UAED: Unsupervised abnormal emotion detection network based on wearable mobile device,” IEEE Transactions on Network Science and Engineering, vol. 10, no. 6, pp. 3682–3696, 2023

    work page 2023

  58. [58]

    Vecaug: Unveiling camouflaged frauds with cohort augmentation for enhanced detection,

    F. Xiao, S. Cai, G. Chen, H. Jagadish, B. C. Ooi, and M. Zhang, “Vecaug: Unveiling camouflaged frauds with cohort augmentation for enhanced detection,” inProceedings of the 30th ACM SIGKDD Confer- ence on Knowledge Discovery and Data Mining, 2024, pp. 6025–6036

    work page 2024

  59. [59]

    Financial fraud:: A review of anomaly detection techniques and recent advances,

    W. Hilal, S. A. Gadsden, and J. Yawney, “Financial fraud:: A review of anomaly detection techniques and recent advances,” 2022

    work page 2022

  60. [60]

    Anomaly detection with adversarial dual autoencoders,

    H. S. Vu, D. Ueta, K. Hashimoto, K. Maeno, S. Pranata, and S. M. Shen, “Anomaly detection with adversarial dual autoencoders,”arXiv preprint arXiv:1902.06924, 2019

    work page arXiv 1902

  61. [61]

    Anomaly detection of time series with smoothness-inducing sequential variational auto-encoder,

    L. Li, J. Yan, H. Wang, and Y . Jin, “Anomaly detection of time series with smoothness-inducing sequential variational auto-encoder,”IEEE transactions on neural networks and learning systems, vol. 32, no. 3, pp. 1177–1191, 2020

    work page 2020

  62. [62]

    Stad-gan: unsupervised anomaly detection on multivariate time series with self-training generative adversarial networks,

    Z. Zhang, W. Li, W. Ding, L. Zhang, Q. Lu, P. Hu, T. Gui, and S. Lu, “Stad-gan: unsupervised anomaly detection on multivariate time series with self-training generative adversarial networks,”ACM Transactions on Knowledge Discovery from Data, vol. 17, no. 5, pp. 1–18, 2023

    work page 2023

  63. [63]

    Imdiffusion: Imputed diffusion models for multivariate time series anomaly detection,

    Y . Chen, C. Zhang, M. Ma, Y . Liu, R. Ding, B. Li, S. He, S. Rajmohan, Q. Lin, and D. Zhang, “Imdiffusion: Imputed diffusion models for multivariate time series anomaly detection,”Proceedings of the VLDB Endowment, vol. 17, no. 3, pp. 359–372, 2023

    work page 2023

  64. [64]

    A diffusion-based framework for multi-class anomaly detection,

    H. He, J. Zhang, H. Chen, X. Chen, Z. Li, X. Chen, Y . Wang, C. Wang, and L. Xie, “A diffusion-based framework for multi-class anomaly detection,” inProceedings of the AAAI conference on artificial intelligence, vol. 38, no. 8, 2024, pp. 8472–8480

    work page 2024

  65. [65]

    Unsuper- vised surface anomaly detection with diffusion probabilistic model,

    X. Zhang, N. Li, J. Li, T. Dai, Y . Jiang, and S.-T. Xia, “Unsuper- vised surface anomaly detection with diffusion probabilistic model,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 6782–6791

    work page 2023

  66. [66]

    Learning from time-changing data with adaptive windowing,

    A. Bifet and R. Gavalda, “Learning from time-changing data with adaptive windowing,” inProceedings of the 2007 SIAM international conference on data mining. SIAM, 2007, pp. 443–448

    work page 2007

  67. [67]

    Fedd: Feature extraction for explicit concept drift detection in time series,

    R. C. Cavalcante, L. L. Minku, and A. L. Oliveira, “Fedd: Feature extraction for explicit concept drift detection in time series,” in2016 International Joint Conference on Neural Networks (IJCNN). IEEE, 2016, pp. 740–747

    work page 2016

  68. [68]

    Detecting volatility shift in data streams,

    D. T. J. Huang, Y . S. Koh, G. Dobbie, and R. Pears, “Detecting volatility shift in data streams,” in2014 IEEE International Conference on Data Mining. IEEE, 2014, pp. 863–868

    work page 2014

  69. [69]

    Dilof: Effective and memory efficient local outlier detection in data streams,

    G. S. Na, D. Kim, and H. Yu, “Dilof: Effective and memory efficient local outlier detection in data streams,” inProceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2018, pp. 1993–2002

    work page 2018

  70. [70]

    Dynamic spatio-temporal graph reasoning for videoqa with self-supervised event recognition,

    J. Nie, X. Wang, R. Hou, G. Li, H. Chen, and W. Zhu, “Dynamic spatio-temporal graph reasoning for videoqa with self-supervised event recognition,”IEEE Transactions on Image Processing, vol. 33, pp. 4145– 4158, 2024. Jiaqi Zhureceived the B.E. degree in automation from the Nanjing Institute of Technology, Nanjing, China, in 2019. She is currently a Ph.D. s...

    work page 2024