Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

CICADA: Cross-Domain Interpretable Coding for Anomaly Detection and Adaptation in Multivariate Time Series

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read CICADA detects anomalies across shifting time-series domains by adapting only one meta-domain per expert at a time.

desk verdict Interesting architecture with a real synthetic proof-of-concept, but the empirical claims are overblown and the single-domain-per-segment assumption is undefended. read the letter →

arxiv 2505.00415 v1 pith:XIKP2EZJ submitted 2025-05-01 cs.LG

Tian Lan , Yifei Gao , Yimeng Lu , Chen Zhang This is my paper · ORCID
classification cs.LG MSC 62M1068T0568T07
keywords anomalydetectionmeta-learningdomainadaptationmixtureofexpertsinterpretabilitynegativetransfermultivariatetimeserieslatentdomains
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes CICADA, an unsupervised method for anomaly detection in multivariate time series whose distribution shifts over time across unknown latent domains. The method trains a mixture of experts, each expert partitioned into several meta-domains, and for each new segment selects only the meta-domain whose parameters fit best, adapting it with a small number of gradient steps. A learnable per-meta-domain step size acts as a signal: if it grows beyond a threshold, a new meta-domain is spawned from the farthest adapted parameters, so newly emerging regimes are absorbed without retraining on old data. A hierarchical attention layer weights the experts by how well they reconstruct each window, giving an interpretable attribution of each expert's contribution. The authors claim that on synthetic, industrial, and public transfer tests this beat existing single-domain and test-time-adaptation anomaly detectors and also identified distinct operating regimes.

What carries the argument

The load-bearing object is the per-expert meta-domain: each expert $j$ partitions the domain space $\mathcal{P}$ into disjoint meta-domains $\mathcal{P}_j^1,\dots,\mathcal{P}_j^{m_j}$, each with its own initial parameters $\Theta_j^k$, and every segment $D_i$ is assigned to exactly one meta-domain per expert through the min-loss indicator $\delta_j^{k\to i}$. The mechanism runs on two signals: the min-selection in Eq. (3), which blocks negative transfer by zeroing gradients from non-selected meta-domains, and the learnable meta-learning rate $\alpha_{\text{meta},j,k}$ in Eq. (2), whose magnitude gauges whether one meta-domain is being stretched across too-dissimilar domains; when it exceeds $h_\alpha$, Eq. (8) triggers expansion. The fusion side uses hierarchical attention: first over meta-domains within an expert, then over experts, with the expert weights in Eq. (10) derived from reconstruction-oriented attention scores. Together these pieces convert an unsupervised detection problem into a sequence of one-step domain selections and occasional domain births.

What would settle it

Take a multivariate series with known ground-truth regime switches, partition it with segment boundaries that intentionally do not align with the switches (for example, segments twice as long as regimes), and run CICADA end to end. If the F1 or AUROC on held-out regime-shifted data does not beat a single-expert, single-meta-domain baseline, or if inspection shows the min-selection in Eq. (3) assigning a segment to a meta-domain that only fits one of its two regimes, the central homogeneity assumption fails in a falsifying way.

Watch

Extended reading notes

Core claim

The central claim is that cross-domain anomaly detection can be recast as per-expert selective adaptation to latent meta-domains, and that this design simultaneously prevents negative transfer, accommodates emerging domains, and yields interpretable expert weights. Formally, for each training segment $D_i$ and each expert $j$, the paper selects the meta-domain $k^* = \arg\min_k \mathcal{L}_j(D_i^{\text{tr}};\Theta_j^k)$ and adapts only from $\Theta_j^{k^*}$, so dissimilar meta-domains are untouched. The learnable meta-learning rate $\alpha_{\text{meta},j,k}$ is interpreted as a measure of within-meta-domain dispersion; when $\alpha_{\text{meta},j,k}>h_\alpha$, a new meta-domain is created with the parameters adapted for the farthest segment, and the old meta-domain is released from that segment. On test segments the same selection and a fixed adaptation step produce anomaly scores, and the final score is a reconstruction-weighted sum over experts via Eq. (10). If this is right, a model can keep growing its domain inventory as regimes appear, without labels, and can tell which expert and which meta-domain is responsible for each detection.

Load-bearing premise

The method assumes every training and test segment is homogeneous: it belongs to exactly one latent domain for its whole duration, so one meta-domain per expert can be selected and adapted for the segment.

Editorial extensions

If this is right

  • If the central claim holds, a deployed detector can handle a stream with regimes that were never seen at training time: the expansion rule adds a meta-domain mid-training, and test segments use only the closest meta-domain, so old regimes are not forgotten when new ones appear.
  • Because each test segment adapts only the selected meta-domain, dissimilar segments do not contaminate each other's adapted parameters; this should reduce false alarms that come from a single shared model averaging incompatible normal behaviors.
  • The expert-attention weights give a per-window, per-expert and per-meta-domain attribution, so an operator can trace an alarm to a particular reconstruction perspective without relying on post-hoc explainers.
  • On the reported transfer experiments, the same hyperparameters transfer across unrelated public datasets, which suggests the method does not need dataset-specific tuning to acquire its cross-domain behavior.

Reading between the lines

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

  • The min-loss selection and the expansion trigger are not tied to anomaly detection; the same mechanism could serve as a general-purpose regime-discovery layer for forecasting or classification, where each expert is a predictive model and each meta-domain is a regime.
  • The paper's homogeneity assumption on segments suggests a testable extension: feed CICADA change-point boundaries from a separate detector, or use overlapping segments, and measure whether F1 improves when regime transitions fall inside segments.
  • The learned $\alpha_{\text{meta},j,k}$ itself could be used as a real-time domain-shift alarm: a spike above threshold after training would signal that the current segment is not well covered by existing meta-domains.
  • Because the expansion initializes a new meta-domain from the farthest adapted parameters, it implicitly assumes new domains resemble some previously encountered domain up to parameter displacement; a completely novel mechanism with no analog in training would require the expert library itself to expand, not just meta-domains.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes CICADA, a mixture-of-experts framework for unsupervised multivariate time series anomaly detection under latent domain shifts. Each expert is associated with several meta-domains, and selective meta-learning adapts one meta-domain per time segment to avoid negative transfer; an adaptive expansion rule grows the number of meta-domains based on learned meta-learning rates, and a hierarchical attention mechanism fuses experts for interpretability. The authors report experiments on synthetic data, three industrial datasets, and transfer experiments on UCR and SMD benchmarks, and claim that CICADA outperforms state-of-the-art baselines in both detection performance and interpretability.

Significance. If the proposed framework worked as claimed, it would address a practically important problem: unsupervised anomaly detection under distributional shifts across multiple latent source and target domains, where domain labels and the number of domains are unknown. The selective meta-learning idea—adapting each expert from only one meta-domain to prevent negative transfer—is original and potentially valuable, and the adaptive expansion mechanism provides a principled way to grow model capacity as new domains appear. The paper gives a detailed algorithmic description (Algorithms 1 and 2) and synthetic experiments that illustrate the expansion mechanism when segment boundaries are known. However, the empirical support for the central claim is undermined by an internal contradiction in the reported results and by an evaluation practice that uses test labels to select the anomaly threshold.

major comments (4)
  1. [Section 3, Algorithm 1, Eq. (3), Eq. (8)] The method assumes every time segment D_i belongs to a single latent domain, but the segmentation procedure is never specified. Algorithm 1 lists 'Partition segments D1,...,DN from X' as a step without giving any method, and the test-time partition is similarly unspecified. The selective meta-learning update (Eq. (3)) and the expansion trigger (Eq. (8)) both rely on segment homogeneity: if a segment straddles two regimes, the min-selection averages incompatible data, the adapted parameters become a compromise, and the claimed protection against negative transfer is lost. Moreover, the expansion trigger uses alpha_meta,j,k, which reflects dispersion across whole segments assigned to a meta-domain, not dispersion inside a mis-segmented segment, so the algorithm cannot detect its own failure. The industrial case studies (Figures 7, 8, 11–13) show detected meta-domains but provide no analysis of segment-boundary accuracy or sensitivity to mis-segmentation. This is a load-bearing gap because the central claim that adapting one meta-domain per segment avoids negative transfer holds only under the homogeneity assumption.
  2. [Section 5.1, Table 1] The text states that 'CICADA significantly outperforms all the other baselines across all the three datasets,' but Table 1 contradicts this. On the Compressor dataset, CICADA's AUROC is 80.6 versus LSTM's 87.3, and its AUPRC is 45.4 versus LSTM's 58.2; on the Busbar dataset, CICADA's AUROC is 77.6 versus M2N2's 85.7. If the claim is meant to apply only to the F1 score, that should be stated explicitly; as written, the claim is false. Since the abstract asserts that CICADA 'outperforms state-of-the-art methods in both cross-domain detection performance and interpretability,' this inconsistency affects the paper's central empirical claim.
  3. [Section 5.2, UCR experiments] The paper says, 'For each method, we use the threshold that maximizes the F1 score to identify anomalies in testing datasets.' Choosing the threshold on the test set using ground-truth labels is a form of fitting to the test data; it inflates reported F1 scores and makes the comparison across methods uninformative, since the reported numbers are not achievable without access to test labels. A proper evaluation should select the threshold on a validation split or report threshold-free metrics such as AUROC, which are already computed elsewhere in the paper. As presented, the UCR results do not support the claim that CICADA 'exhibits robust generalization performance.'
  4. [Section 3.3, Eq. (8)] The expansion condition asserts that a large learned meta-learning rate alpha_meta,j,k indicates that the domains within a meta-domain are heterogeneous and that a new meta-domain is needed. This link is not justified. alpha_meta,j,k is a single scalar per meta-domain, regularized toward small values by the penalty term L_pen,j in Eq. (5), and it is learned by gradient descent; the paper provides no theoretical argument or experimental evidence that the optimal alpha reflects within-meta-domain dispersion. The synthetic experiments (Section 4.2) use known segment boundaries and four clearly separated domains, and they do not test gradual shifts or mis-specified boundaries, where the proposed signal would likely fail.
minor comments (6)
  1. [Throughout] There are repeated typos, e.g., 'heterogenous' for 'heterogeneous' and 'time-time adaption' for 'test-time adaptation' in Section 5.1; the manuscript should be proofread.
  2. [Section 3.2, after Eq. (4)] The sentence 'To ensure that the initial parameters Theta^k_j of each meta-domain are closed to the corresponding adapted parameters of each domain' should read 'close to' rather than 'closed to'.
  3. [Section 3.4, Eq. (10)] The attention weights in Eq. (10) are defined using only the query and key; the value is ignored in the weighting. Please clarify whether the value features are used for the reconstruction and whether the weight w_j(X_t) is intended to be derived from attention scores only, as the notation suggests.
  4. [Section 5.1, ablation] The ablation CICADAna is described only as 'not adding' the adaptive expansion mechanism; the description does not state whether the number of meta-domains is fixed to one, which meta-learning rates are used, or how the ablation is trained. More details are needed to interpret the comparison.
  5. [Figures 4–6] The experimental figures report curves without error bars or repeated-run variance. Given the stochastic nature of the training process, error bars or at least a statement about the number of runs would strengthen the interpretability and expansion results.
  6. [Section 2] The related work claims that 'no work has specifically addressed domain adaptation for time-varying and latent domains,' but the paper itself cites test-time adaptation methods (e.g., M2N2) and several domain-adaptation methods for time series; the claim as written is too strong and should be qualified.

Circularity Check

1 steps flagged · score 2.0 of 10

No derivation-level circularity; the only fitting-to-data is the oracle UCR threshold, which slightly inflates reported F1 but does not infect the learning chain.

  1. fitted input called prediction [Section 5.2, UCR transferability experiments]
    "For each method, we use the threshold that maximizes the F1 score to identify anomalies in testing datasets."

    The anomaly labels of the UCR test sets are used to select the decision threshold C; the reported F1 is then by construction the maximum F1 attainable for that anomaly-score vector, rather than the performance of a threshold chosen from training data alone. The headline UCR performance numbers are therefore fitted values, and the claim that CICADA 'outperforms' on those datasets is partly a statement about a test-set-optimized decision rule rather than an independent prediction. This does not make the meta-domain learning mechanism itself circular, since the anomaly scores are computed without test labels, so this is a partial evaluation-protocol fitting rather than a derivation-level circularity.

full rationale

The paper's central derivation chain is self-contained. The four claimed innovations are defined operationally: selective meta-learning selects the meta-domain with minimum loss (Eq. 3), adaptive expansion triggers when a learned meta-learning rate exceeds a hand-set threshold (Eq. 8), and expert weights come from reconstruction-based attention (Eq. 10). These are algorithmic operations, not predictions derived from fitted constants, and no equation in the paper reduces to another by construction. There are no self-citations: the reference list contains no work by the present authors, so no load-bearing argument is imported from the authors' own prior work, and no uniqueness theorem is borrowed. The homogeneity assumption that each segment belongs to one latent domain is a correctness risk if segment boundaries straddle regimes, but it is an explicit modeling assumption, not a circular step. The main weakness is the UCR evaluation protocol, where the threshold is selected on test labels to maximize F1; this makes the reported F1 a fitted maximum for each score vector. Because the same protocol is applied to all baselines and because the anomaly scores themselves are free of test labels, the fitting is a limited evaluation flaw rather than a proof that the method's claims reduce to their inputs. Overall score 2 reflects this one evaluation-fitting practice, with no circularity in the learning mechanism, the expansion rule, or the interpretability analysis.

Assumptions & free parameters 8 free parameters · 6 assumptions · 1 invented entities

The central method rests on a small set of hand-set hyperparameters and on the latent-domain homogeneity assumption. The only invented construct is the meta-domain. The heaviest unstated input is the test-set threshold choice, which affects all reported F1 numbers.

free parameters (8)
  • Expert count m = 5 synthetic, 4 or 3 industrial, 3 public
    Hand-selected; no criterion for choosing the expert set or number of experts is given.
  • Expansion threshold h_alpha = 5e-4 for public experiments; not stated for industrial
    Controls when Eq (8) adds a meta-domain; the number of discovered domains and the reported performance depend on this threshold.
  • Meta-learning rate alpha_meta,j,k = learned per meta-domain
    The magnitude of this learned rate is the expansion signal in Eq (8), so the domain-discovery behavior depends on its optimization trajectory and penalty.
  • Test-time adaptation rate alpha_1 = 1e-3 in public experiments
    Fixed step for adapting meta-domains to test segments in Section 3.6; no sensitivity analysis is provided.
  • Loss balance lambda_1 = 1000 in public experiments
    Balances reconstruction and extraction losses in Eq (12); industrial values are not stated.
  • Meta-learning rate penalty lambda_pen = not specified
    Appears in Eq (5) as the weight on L_pen; no numeric value is given.
  • Anomaly threshold C = maximizes F1 on test sets for UCR; 99.5th percentile for SMD
    In Section 5.2 the threshold is chosen on the testing data, which makes reported F1 an in-sample optimum rather than a predictive estimate.
  • Window size L = 1, 10, or 30 depending on experiment
    Handpicked per dataset in Appendix C and Section 5.2; no principled selection procedure is given.
assumptions (6)
  • domain assumption Each time segment D_i belongs to exactly one latent domain, so one meta-domain can be selected for the whole segment.
    Section 3 problem formulation and Algorithm 1 partition the series before domain selection; mixed segments break min-selection in Eq (3).
  • domain assumption The meta-domains P_j^k form a disjoint partition of the bounded domain space P for every expert.
    Section 3.1 states that P_j^k are disjoint and union to P, but no justification is given that each expert's division is coherent or stable.
  • domain assumption The meta-domain with the smallest training loss is the correct one to adapt for a segment.
    Eq (3) defines L_domain as the min over k and Eq (4) uses the argmin indicator; under noisy or uninformative losses the lowest-loss domain may be a poor match for anomaly scoring.
  • domain assumption A single gradient step with the learned alpha adequately adapts a meta-domain to a segment.
    Eq (2) applies one first-order MAML-style step; the paper provides no convergence or robustness analysis.
  • domain assumption Reconstruction quality is a valid proxy for expert weight in anomaly detection.
    Section 3.4 weights experts by their reconstruction contribution; reconstruction error and anomaly score do not necessarily rank experts identically.
  • standard math Gradient-based optimization of the specified statistical and neural expert losses is valid.
    The method relies on projected gradient descent, QR decomposition, Kernel Hebbian updates, and backpropagation; these are standard techniques.
invented entities (1)
  • Meta-domain (P_j^k, Theta_j^k)
    purpose: A per-expert partition of the domain space with an initial parameter set; the algorithm selects one meta-domain per segment for adaptation and can split it when alpha exceeds h_alpha.
    No external falsifiable handle is given; the construct is validated only by downstream performance on synthetic and industrial data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CICADA: Cross-Domain Interpretable Coding for Anomaly Detection and Adaptation in Multivariate Time Series." pith.science (2026). https://pith.science/paper/XIKP2EZJ

@misc{pith2026250500415,
  author       = {Pith},
  title        = {Pith review of: CICADA: Cross-Domain Interpretable Coding for Anomaly Detection and Adaptation in Multivariate Time Series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XIKP2EZJ}},
  note         = {Machine review of arXiv:2505.00415}
}
read the original abstract

Unsupervised Time series anomaly detection plays a crucial role in applications across industries. However, existing methods face significant challenges due to data distributional shifts across different domains, which are exacerbated by the non-stationarity of time series over time. Existing models fail to generalize under multiple heterogeneous source domains and emerging unseen new target domains. To fill the research gap, we introduce CICADA (Cross-domain Interpretable Coding for Anomaly Detection and Adaptation), with four key innovations: (1) a mixture of experts (MOE) framework that captures domain-agnostic anomaly features with high flexibility and interpretability; (2) a novel selective meta-learning mechanism to prevent negative transfer between dissimilar domains, (3) an adaptive expansion algorithm for emerging heterogeneous domain expansion, and (4) a hierarchical attention structure that quantifies expert contributions during fusion to enhance interpretability further.Extensive experiments on synthetic and real-world industrial datasets demonstrate that CICADA outperforms state-of-the-art methods in both cross-domain detection performance and interpretability.

Figures

Figures reproduced from arXiv: 2505.00415 by the authors.

Figure 1
Figure 1. Cross-domain Anomaly Detection Problem single source and a single target domain. Few studies have ad￾dressed multi-source domain adaptation [37] or generalization [8], which aim to leverage data from all the source domains to extract invariant features applicable to the target domain. However, this approach does not account for the potential heterogeneity between domains. Constructing a common feature space for hete… view at source ↗
Figure 2
Figure 2. The framework of CICADA experts. Hence for a domain P𝑑𝑖 , its meta-domain label can be given by [𝑘1, 𝑘2, . . . , 𝑘𝑚]. To detect anomalous observations in X′ , for any X′ 𝑡 ∈ D′ 𝑖 ′ , we first get its adapted model parameters Θ 𝑘→𝑖 ′ 𝑗 , and then extract the corresponding extracted feature as follows: f 𝑘 ′ 𝑗,𝑡 = 𝑔𝑗 (X′ 𝑡 ; Θ 𝑘→𝑖 ′ 𝑗 ), ∀X′ 𝑡 ∈ D′ 𝑖 ′ . We also define another anomaly score function 𝐴𝑗 (X′ 𝑡 ; Θ 𝑘→𝑖 ′… view at source ↗
Figure 3
Figure 3. (a): Time series with multiple domains; (b): The idea [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (8 more)
Figure 6
Figure 6. Figure 6: Meta-learning rate and meta-domain assignment [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: Different meta-domains detected by expert SFA for [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: F1 score, AUROC and AUPRC for UCR Datasets [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: F1 score, precision and recall for SMD Datasets [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: Different meta-domains detected by expert PCA for the FluorinePump Dataset [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]
Figure 12
Figure 12. Figure 12: Different meta-domains detected by expert PCA for the Busbar Dataset [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]
Figure 13
Figure 13. Figure 13: Different meta-domains detected by expert SFA for the Busbar Dataset [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]
Figure 14
Figure 14. Figure 14: Evolutionary insight for selective meta-learning and adaptive expansion [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Towards Foundation Models for Zero-Shot Time Series Anomaly Detection: Leveraging Synthetic Data and Relative Context Discrepancy

    cs.LG 2025-09 conditional novelty 6.0 of 10

    TimeRCD, a transformer pre-trained on 2.5 billion synthetic data points with context-relative anomaly labels, beats reconstruction-based foundation models on most zero-shot TSAD benchmarks, while the evaluation tunes ...

Reference graph

Works this paper leans on

51 extracted references · 24 canonical work pages · cited by 1 Pith paper

  1. [1]

    Julien Audibert, Pietro Michiardi, Frédéric Guyard, Sébastien Marti, and Maria A Zuluaga. 2020. Usad: Unsupervised anomaly detection on multivariate time series. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining . 3395–3404

    2020

  2. [2]

    Ruichu Cai, Jiawei Chen, Zijian Li, Wei Chen, Keli Zhang, Junjian Ye, Zhuozhang Li, Xiaoyan Yang, and Zhenjie Zhang. 2021. Time series domain adaptation via sparse associative structure alignment. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 35. 6859–6867

    2021

  3. [3]

    Ziliang Chen, Jingyu Zhuang, Xiaodan Liang, and Liang Lin. 2019. Blending-target domain adaptation by adversarial meta-adaptation networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2248–2257

    2019

  4. [4]

    Giordano Colò. 2019. Anomaly detection for cyber security: time series forecast- ing and deep learning. (2019)

    work page 2019

  5. [5]

    Zahra Zamanzadeh Darban, Geoffrey I Webb, Shirui Pan, Charu C Aggarwal, and Mahsa Salehi. 2025. CARLA: Self-supervised contrastive representation learning for time series anomaly detection. Pattern Recognition 157 (2025), 110874

    2025

  6. [6]

    Zahra Zamanzadeh Darban, Yiyuan Yang, Geoffrey I Webb, Charu C Aggar- wal, Qingsong Wen, and Mahsa Salehi. 2024. DACAD: Domain adaptation con- trastive learning for anomaly detection in multivariate time series. arXiv preprint arXiv:2404.11269 (2024)

    arXiv 2024

  7. [7]

    Hoang Anh Dau, Anthony Bagnall, Kaveh Kamgar, Chin-Chia Michael Yeh, Yan Zhu, Shaghayegh Gharghabi, Chotirat Ann Ratanamahatana, and Eamonn Keogh

  8. [8]

    Songgaojun Deng, Olivier Sprangers, Ming Li, Sebastian Schelter, and Maarten de Rijke. 2024. Domain Generalization in Time Series Forecasting.ACM Transactions on Knowledge Discovery from Data 18, 5 (2024), 1–24

    work page 2024

Show all 51 references
  1. [9]

    Yong Feng, Jinglong Chen, Zhuozheng Yang, Xiaogang Song, Yuanhong Chang, Shuilong He, Enyong Xu, and Zitong Zhou. 2021. Similarity-based meta-learning network with adversarial domain adaptation for cross-domain fault identification. Knowledge-Based Systems 217 (2021), 106829

    2021

  2. [10]

    Chris Fifty, Ehsan Amid, Zhe Zhao, Tianhe Yu, Rohan Anil, and Chelsea Finn

  3. [11]

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. 2017. Model-agnostic meta- learning for fast adaptation of deep networks. In International conference on machine learning. PMLR, 1126–1135

    2017

  4. [12]

    Siho Han and Simon S Woo. 2022. Learning sparse latent graph representations for anomaly detection in multivariate time series. In Proceedings of the 28th ACM SIGKDD Conference on knowledge discovery and data mining . 2977–2986

    2022

  5. [13]

    Huan He, Owen Queen, Teddy Koker, Consuelo Cuevas, Theodoros Tsiligkaridis, and Marinka Zitnik. 2023. Domain adaptation for time series under feature and label shifts. InInternational Conference on Machine Learning. PMLR, 12746–12774

    2023

  6. [14]

    Yifan He, Yatao Bian, Xi Ding, Bingzhe Wu, Jihong Guan, Ji Zhang, and Shuigeng Zhou. 2024. Variate Associated Domain Adaptation for Unsupervised Multivariate Time Series Anomaly Detection. ACM Transactions on Knowledge Discovery from Data (2024)

    2024

  7. [15]

    Xiaoyong Jin, Youngsuk Park, Danielle Maddix, Hao Wang, and Yuyang Wang

  8. [16]

    Keogh, T

    E. Keogh, T. Dutta Roy, U. Naik, and A. Agrawal. 2021. Multi-dataset Time-Series Anomaly Detection Competition. In SIGKDD 2021. https://compete.hexagon- ml.com/practice/competition/39/

    2021

  9. [17]

    Dongmin Kim, Sunghyun Park, and Jaegul Choo. 2024. 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. 13113–13121

    2024

  10. [18]

    Kwang In Kim, M Franz, and Bernhard Schölkopf. 2003. Kernel Hebbian algorithm for iterative kernel principal component analysis. (2003)

    2003

  11. [19]

    Kwei-Herng Lai, Lan Wang, Huiyuan Chen, Kaixiong Zhou, Fei Wang, Hao Yang, and Xia Hu. 2023. Context-aware domain adaptation for time series anomaly detection. In Proceedings of the 2023 SIAM International Conference on Data Mining (SDM). SIAM, 676–684

    2023

  12. [20]

    Da Li, Yongxin Yang, Yi-Zhe Song, and Timothy Hospedales. 2018. Learning to generalize: Meta-learning for domain generalization. In Proceedings of the AAAI conference on artificial intelligence , Vol. 32

    2018

  13. [21]

    Hongbo Li, Wenli Zheng, Feilong Tang, Yanmin Zhu, and Jielong Huang. 2023. Few-shot time-series anomaly detection with unsupervised domain adaptation. Information Sciences 649 (2023), 119610

    2023

  14. [22]

    Yuxin Li, Wenchao Chen, Bo Chen, Dongsheng Wang, Long Tian, and Mingyuan Zhou. 2023. Prototype-oriented unsupervised anomaly detection for multivariate time series. InInternational Conference on Machine Learning. PMLR, 19407–19424

    2023

  15. [23]

    Yiying Li, Yongxin Yang, Wei Zhou, and Timothy Hospedales. 2019. Feature-critic networks for heterogeneous domain generalization. In International Conference on Machine Learning. PMLR, 3915–3924

    2019

  16. [24]

    Zhihan Li, Youjian Zhao, Jiaqi Han, Ya Su, Rui Jiao, Xidao Wen, and Dan Pei. 2021. 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 & da...

    2021

  17. [25]

    Yajing Liu, Yuning Lu, Hao Liu, Yaozu An, Zhuoran Xu, Zhuokun Yao, Baofeng Zhang, Zhiwei Xiong, and Chenguang Gui. 2023. Hierarchical prompt learning for multi-task learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 10888–10898

    2023

  18. [26]

    Wang Lu, Jindong Wang, Xinwei Sun, Yiqiang Chen, and Xing Xie. 2022. Out-of- distribution representation learning for time series classification. arXiv preprint arXiv:2209.07027 (2022)

    2022 arXiv

  19. [27]

    Pankaj Malhotra, Anusha Ramakrishnan, Gaurangi Anand, Lovekesh Vig, Puneet Agarwal, and Gautam Shroff. 2016. LSTM-based encoder-decoder for multi-sensor anomaly detection. arXiv preprint arXiv:1607.00148 (2016)

    2016 arXiv

  20. [28]

    Felix Ott, David Rügamer, Lucas Heublein, Bernd Bischl, and Christopher Mutschler. 2022. Domain adaptation for time-series classification to mitigate covariate shift. In Proceedings of the 30th ACM international conference on multi- media. 5934–5943

    2022

  21. [29]

    Joao Pereira and Margarida Silveira. 2019. Learning representations from health- care time series data for unsupervised anomaly detection. In 2019 IEEE interna- tional conference on big data and smart computing (BigComp) . IEEE, 1–7

    2019

  22. [30]

    Lifeng Shen, Zhuocong Li, and James Kwok. 2020. Timeseries anomaly detection using temporal hierarchical one-class network. Advances in Neural Information Processing Systems 33 (2020), 13016–13026

    2020

  23. [31]

    Yang Shu, Zhangjie Cao, Chenyu Wang, Jianmin Wang, and Mingsheng Long

  24. [32]

    Xiaozhuang Song, Shun Zheng, Wei Cao, James Yu, and Jiang Bian. 2022. Effi- cient and effective multi-task grouping via meta learning on task combinations. Advances in Neural Information Processing Systems 35 (2022), 37647–37659

    2022

  25. [33]

    Trevor Standley, Amir Zamir, Dawn Chen, Leonidas Guibas, Jitendra Malik, and Silvio Savarese. 2020. Which tasks should be learned together in multi-task learning?. In International conference on machine learning . PMLR, 9120–9132

    2020

  26. [34]

    Ya Su, Youjian Zhao, Chenhao Niu, Rong Liu, Wei Sun, and Dan Pei. 2019. Robust anomaly detection for multivariate time series through stochastic recurrent neural network. In Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining . 2828–2837

    2019

  27. [35]

    In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Open domain generalization with domain-augmented meta-learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 9624–9633

  28. [36]

    Rui Wang, Chongwei Liu, Xudong Mou, Kai Gao, Xiaohui Guo, Pin Liu, Tianyu Wo, and Xudong Liu. 2023. Deep contrastive one-class time series anomaly detection. In Proceedings of the 2023 SIAM International Conference on Data Mining (SDM). SIAM, 694–702

    2023

  29. [37]

    Garrett Wilson, Janardhan Rao Doppa, and Diane J Cook. 2023. CALDA: Im- proving multi-source time series domain adaptation with contrastive adversarial learning. IEEE transactions on pattern analysis and machine intelligence (2023)

    2023

  30. [38]

    Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. 2022. Timesnet: Temporal 2d-variation modeling for general time series analysis. arXiv preprint arXiv:2210.02186 (2022)

    2022 arXiv

  31. [39]

    Franco Van Wyk, Yiyang Wang, Anahita Khojandi, and Neda Masoud. 2019. Real- time sensor anomaly detection and identification in automated vehicles. IEEE Transactions on Intelligent Transportation Systems 21, 3 (2019), 1264–1276

    2019

  32. [40]

    Jiehui Xu. 2021. Anomaly transformer: Time series anomaly detection with association discrepancy. arXiv preprint arXiv:2110.02642 (2021)

    2021 arXiv

  33. [41]

    Yiyuan Yang, Chaoli Zhang, Tian Zhou, Qingsong Wen, and Liang Sun. 2023. Dcdetector: Dual attention contrastive representation learning for time series anomaly detection. In Proceedings of the 29th ACM SIGKDD Conference on Knowl- edge Discovery and Data Mining . 3033–3045

    2023

  34. [42]

    Huaxiu Yao, Ying Wei, Junzhou Huang, and Zhenhui Li. 2019. Hierarchically structured meta-learning. In International conference on machine learning . PMLR, 7045–7054

    2019

  35. [43]

    Hongzuo Xu, Yijie Wang, Songlei Jian, Qing Liao, Yongjun Wang, and Guansong Pang. 2024. Calibrated one-class classification for unsupervised time series anomaly detection. IEEE Transactions on Knowledge and Data Engineering (2024)

    2024

  36. [44]

    Wenyu Zhang, Mohamed Ragab, and Chuan-Sheng Foo. 2022. Domain generaliza- tion via selective consistency regularization for time series classification. In 2022 26th International Conference on Pattern Recognition (ICPR) . IEEE, 2149–2156

    2022

  37. [45]

    Yuxin Zhang, Yiqiang Chen, Jindong Wang, and Zhiwen Pan. 2021. Unsupervised deep anomaly detection for multi-sensor time-series signals. IEEE Transactions on Knowledge and Data Engineering 35, 2 (2021), 2118–2132

    2021

  38. [46]

    Sicheng Zhao, Hui Chen, Hu Huang, Pengfei Xu, and Guiguang Ding. 2024. More is Better: Deep Domain Adaptation with Multiple Sources. arXiv preprint arXiv:2405.00749 (2024)

    2024 arXiv

  39. [47]

    Zahra Zamanzadeh Darban, Geoffrey I Webb, Shirui Pan, Charu Aggarwal, and Mahsa Salehi. 2024. Deep learning for time series anomaly detection: A survey. Comput. Surveys 57, 1 (2024), 1–42

    2024

  40. [51]

    Yue Zhao, Ryan Rossi, and Leman Akoglu. 2021. Automatic unsupervised outlier model selection. Advances in Neural Information Processing Systems 34 (2021), 4489–4502. CICADA: Cross-Domain Interpretable Coding for Anomaly Detection and Adaptation in Multivariate Time Series Conf...

    2021

  41. [2019]

    IEEE/CAA Journal of Automatica Sinica 6, 6 (2019), 1293–1305

    The UCR time series archive. IEEE/CAA Journal of Automatica Sinica 6, 6 (2019), 1293–1305

    2019

  42. [2021]

    Advances in Neural Information Processing Systems 34 (2021), 27503–27516

    Efficiently identifying task groupings for multi-task learning. Advances in Neural Information Processing Systems 34 (2021), 27503–27516

    2021

  43. [2022]

    In International Conference on Machine Learning

    Domain adaptation for time series forecasting via attention sharing. In International Conference on Machine Learning . PMLR, 10280–10297

Pith tools

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