pith. sign in

arxiv: 2503.13868 · v3 · pith:UJ5QINYAnew · submitted 2025-03-18 · 💻 cs.LG · cs.AI

Out-of-Distribution Generalization in Time Series: A Survey

Xin Wu , Fei Teng , Xingwang Li , Ji Zhang , Tianrui Li , Qiang Duan This is my paper

Pith reviewed 2026-05-23 00:13 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords out-of-distribution generalizationtime seriessurveydata distributionrepresentation learningOOD evaluationdistribution shiftsmachine learning
0
0 comments X

The pith

The first survey organizes out-of-distribution generalization methods for time series across data distribution, representation learning, and OOD evaluation.

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

Time series data in open environments frequently encounter distribution shifts, latent feature diversity, and non-stationary dynamics that degrade standard model performance. This survey synthesizes the literature to supply the first structured map of methodologies aimed at achieving out-of-distribution generalization in this setting. It partitions the work into three dimensions to trace development and current approaches, detailing algorithms within each, noting applications, and listing open challenges. A reader would care because such a map reduces the effort needed to locate relevant techniques and to recognize where progress remains limited.

Core claim

The paper claims to deliver the first comprehensive review of OOD generalization methodologies for time series. It organizes the analysis across three foundational dimensions—data distribution, representation learning, and OOD evaluation—and presents several popular algorithms in detail for each. The review further highlights key application scenarios, identifies persistent challenges, and proposes future research directions while providing an online summary of the covered methods.

What carries the argument

Three foundational dimensions—data distribution, representation learning, and OOD evaluation—that organize the review of algorithms and delineate the field's trajectory.

If this is right

  • Methods become comparable within each dimension, revealing patterns in how distribution shifts are handled.
  • Application scenarios receive explicit emphasis, connecting techniques to forecasting, anomaly detection, and similar tasks.
  • Persistent challenges are isolated so that subsequent work can target them directly.
  • An online table of reviewed methods supplies a reference that accelerates lookup and gap identification.

Where Pith is reading between the lines

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

  • The three-dimension structure could be tested for extension to other sequential data types that share non-stationarity.
  • Standardized benchmarks might emerge from the OOD evaluation dimension to compare methods more consistently across papers.
  • Future updates to the survey could track whether new algorithms continue to cluster inside the existing dimensions or require an added category.

Load-bearing premise

The three chosen dimensions together with the selected popular algorithms provide a complete and unbiased coverage of the OOD generalization literature for time series.

What would settle it

Discovery of a significant OOD generalization method for time series that cannot be placed in any of the three dimensions, or omission of widely used algorithms from the detailed presentations, would show the framework incomplete.

Figures

Figures reproduced from arXiv: 2503.13868 by Fei Teng, Ji Zhang, Qiang Duan, Tianrui Li, Xingwang Li, Xin Wu.

Figure 1
Figure 1. Figure 1: Examples illustrating how real-world temporal dynamics drive distribution shifts in TS-OOG. The first two columns show covariate shift, where the input feature distribution (ℙ(𝑋)) changes due to underlying dynamics like gradual temporal drifts (e.g., user growth) or abrupt events (e.g., server upgrade). The third column depicts concept drift, a more fundamental change where the input-output relationship (ℙ… view at source ↗
Figure 2
Figure 2. Figure 2: A research roadmap of representative methods for TS-OOG. The methods are categorized along three dimensions: data distribution assumptions (indicated by border color), representation learning strategies (indicated by line style), and OOD evaluation protocols (indicated by block color). Methods within the roadmap are presented in chronological order. For a more comprehensive summary, please refer to the mai… view at source ↗
Figure 3
Figure 3. Figure 3: A comprehensive taxonomy of TS-OOG methods. and future directions, respectively. Finally, section 9 con￾cludes the survey. X. Wu et al.: Preprint submitted to Elsevier Page 3 of 21 [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: A taxonomy examining methods for TS-OOG from three perspectives: data distribution, representation learning, and OOD evaluation. This includes various approaches like covariate shift adaptation, invariant learning, adaptive mechanism-based methods, and large time series models (LTSMs). Note that OOD evaluation specifically for LTSMs is not summarized here due to the current lack of standardized evaluation … view at source ↗
Figure 5
Figure 5. Figure 5: Each circle represents a data instance, with different colors indicating the class to which the instance belongs. (a). Original data: The distribution of instances and the classification boundary (dashed line) remain stable. (b). Co￾variate shift: The distribution of instances changes (the feature distribution ℙ(𝑋) changes), but the class decision boundary ℙ(𝑌 |𝑋) remains unchanged. (c). Concept drift: The… view at source ↗
Figure 6
Figure 6. Figure 6: A taxonomy of LTSMs. LLM-based methods: These methods leverage pre-trained LLMs. Time series data is first converted into a token format that LLMs can understand, via techniques like patching (for tuning-based methods) or textual prompts (for prompt-based methods). The LLM is then used to perform the time series task without being retrained from scratch. Time series foundation models: These are foundation … view at source ↗
Figure 7
Figure 7. Figure 7: Several applications of TS-OOG. effectively integrated via TS-OOG techniques, can sub￾stantially improve the intelligence of urban management [36, 174, 158]. In transportation, demand for data distri￾bution–agnostic solutions extends to real-time decision￾making and large-scale data integration and management. 2) Environment. Air quality data often display dynamic characteristics influenced by seasonal var… view at source ↗
read the original abstract

Time series frequently manifest distribution shifts, diverse latent features, and non-stationary learning dynamics, particularly in open and evolving environments. These characteristics pose significant challenges for out-of-distribution (OOD) generalization. While substantial progress has been made, a systematic synthesis of advancements remains lacking. To address this gap, we present the first comprehensive review of OOD generalization methodologies for time series, organized to delineate the field's evolutionary trajectory and contemporary research landscape. We organize our analysis across three foundational dimensions: data distribution, representation learning, and OOD evaluation. For each dimension, we present several popular algorithms in detail. Furthermore, we highlight key application scenarios, emphasizing their real-world impact. Finally, we identify persistent challenges and propose future research directions. A detailed summary of the methods reviewed for the generalization of OOD in time series can be accessed at https://tsood-generalization.com.

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

3 major / 2 minor

Summary. The manuscript is a survey claiming to be the first comprehensive review of out-of-distribution (OOD) generalization methodologies for time series. It organizes the literature along three dimensions—data distribution, representation learning, and OOD evaluation—detailing popular algorithms in each, highlighting application scenarios, identifying challenges, and proposing future directions, with a supplementary website providing a detailed summary of reviewed methods.

Significance. If the three-dimensional taxonomy provides balanced coverage without major omissions, the survey would offer a useful synthesis of an evolving field, delineating its trajectory and landscape for researchers. The provision of a public website for method summaries strengthens accessibility and could aid reproducibility of the literature mapping.

major comments (3)
  1. [Introduction] Introduction: The assertion that this is the 'first comprehensive review' requires explicit comparison to prior surveys on OOD generalization or time-series robustness (e.g., any existing reviews on domain adaptation or non-stationary time series) to substantiate novelty and completeness; without this, the central claim risks being overstated.
  2. [§3] §3 (or equivalent section on the three dimensions): The choice of data distribution, representation learning, and OOD evaluation as the 'foundational dimensions' lacks a clear justification or mapping to why these axes together capture the full space of time-series OOD methods; this directly affects whether the organization delineates the 'evolutionary trajectory' as claimed.
  3. [algorithm sections] Sections detailing algorithms per dimension: Criteria for selecting 'popular algorithms' are not stated, raising the possibility of selection bias; for the survey to support its claim of comprehensive coverage, explicit inclusion/exclusion rules (e.g., citation count, recency, or impact) should be provided.
minor comments (2)
  1. [Abstract/Introduction] The abstract and introduction could more precisely define what constitutes an 'OOD' shift in the time-series context (e.g., covariate vs. concept shift) to ground the subsequent taxonomy.
  2. [throughout] Ensure all referenced methods on the supplementary website are cross-cited in the main text with consistent notation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We appreciate the referee's constructive feedback and recommendation for minor revision. We address each major comment point by point below.

read point-by-point responses

  1. Referee: [Introduction] Introduction: The assertion that this is the 'first comprehensive review' requires explicit comparison to prior surveys on OOD generalization or time-series robustness (e.g., any existing reviews on domain adaptation or non-stationary time series) to substantiate novelty and completeness; without this, the central claim risks being overstated.

    Authors: We agree that an explicit comparison to prior surveys would strengthen the claim of novelty. While no comprehensive survey exists specifically on OOD generalization for time series, we will add a dedicated paragraph (and possibly a comparison table) in the Introduction section that contrasts our work with related surveys on domain adaptation for time series, non-stationary time series, and general OOD methods in other modalities. This revision will better substantiate completeness. revision: yes

  2. Referee: [§3] §3 (or equivalent section on the three dimensions): The choice of data distribution, representation learning, and OOD evaluation as the 'foundational dimensions' lacks a clear justification or mapping to why these axes together capture the full space of time-series OOD methods; this directly affects whether the organization delineates the 'evolutionary trajectory' as claimed.

    Authors: The three dimensions were chosen because they align with the core stages of OOD handling in time series (data-level shift mitigation, invariant feature learning, and rigorous evaluation). We will revise the relevant section (currently §3) to provide an explicit justification and mapping, including how the axes together cover the methodological landscape and trace evolutionary trends. A clarifying diagram may also be added. revision: yes

  3. Referee: [algorithm sections] Sections detailing algorithms per dimension: Criteria for selecting 'popular algorithms' are not stated, raising the possibility of selection bias; for the survey to support its claim of comprehensive coverage, explicit inclusion/exclusion rules (e.g., citation count, recency, or impact) should be provided.

    Authors: We agree that stating selection criteria is necessary to address potential bias concerns. In the revised manuscript we will add an explicit paragraph (or short 'Scope and Selection' subsection) at the start of the algorithm sections, specifying that we prioritized highly cited works, recent high-impact papers (primarily post-2020) with demonstrated relevance to time-series OOD, and representative methods across sub-areas, while noting the goal of illustrative rather than exhaustive coverage. revision: yes

Circularity Check

0 steps flagged

No significant circularity in survey synthesis

full rationale

This is a literature survey paper whose central contribution is a taxonomic organization of existing external work across three dimensions (data distribution, representation learning, OOD evaluation). No derivations, equations, predictions, or fitted parameters are introduced that could reduce to quantities defined inside the paper. All algorithms and results discussed are drawn from prior literature; the paper performs synthesis rather than any self-referential computation or uniqueness proof. The claim of providing the 'first comprehensive review' is a statement of scope, not a load-bearing logical step that collapses by construction. Self-citations, if present, are not used to justify any internal result.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the authors' selection of literature and the validity of their proposed three-dimensional taxonomy; these are not independently verified from the abstract alone.

axioms (1)
  • domain assumption The OOD generalization literature for time series can be meaningfully partitioned into the three dimensions of data distribution, representation learning, and OOD evaluation.
    This partitioning is presented as the organizational backbone in the abstract.

pith-pipeline@v0.9.0 · 5684 in / 1112 out tokens · 68258 ms · 2026-05-23T00:13:23.084563+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 3 Pith papers

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

  1. Assessing the Robustness of Climate Foundation Models under No-Analog Distribution Shifts

    cs.LG 2026-03 conditional novelty 5.0

    Climate foundation models restricted to historical-only training exhibit an accuracy-stability trade-off under no-analog forcing shifts, with the ClimaX model showing the lowest absolute errors but up to 8.44% relativ...

  2. ERIS: An Energy-Guided Feature Disentanglement Framework for Out-of-Distribution Time Series Classification

    cs.LG 2025-08 unverdicted novelty 5.0

    ERIS combines energy-guided calibration, weight-level orthogonality, and auxiliary adversarial generalization to produce shift-robust representations for out-of-distribution time series classification.

  3. Cross-Machine Anomaly Detection Leveraging Pre-trained Time-series Model

    cs.LG 2026-04 unverdicted novelty 4.0

    A cross-machine anomaly detection framework disentangles MOMENT embeddings using random forests to create machine-invariant condition features that improve generalization to unseen machines on industrial data.

Reference graph

Works this paper leans on

194 extracted references · 194 canonical work pages · cited by 3 Pith papers · 4 internal anchors

  1. [1]

    ImmFusion: Robust mmWave-RGB Fusion for 3D Human Body Reconstruction in All Weather Conditions

    Agarwal, S., Fridovich-Keil, D., Chinchali, S.P., 2023. Robust forecasting for robotic control: A game-theoretic approach, in: Pro- ceedingsofthe40thIEEEInternationalConferenceonRoboticsand Automation, pp. 5566–5573. doi:10.1109/ICRA48891.2023.10160721

    work page doi:10.1109/icra48891.2023.10160721 2023

  2. [2]

    Advances in diffusion models for image data augmentation: A review of methods, models, evaluation metricsandfutureresearchdirections

    Alimisis,P.,Mademlis,I.,Radoglou-Grammatikis,P.,Sarigiannidis, P., Papadopoulos, G.T., 2025. Advances in diffusion models for image data augmentation: A review of methods, models, evaluation metricsandfutureresearchdirections. ArtificialIntelligenceReview 58, 112. doi:10.1007/s10462-025-11116-x

    work page doi:10.1007/s10462-025-11116-x 2025

  3. [3]

    Chronos: Learning the language of time series

    Ansari,A.F.,Stella,L.,Turkmen,C.,Zhang,X.,Mercado,P.,Shen, H., Shchur, O., Rangapuram, S.S., Pineda Arango, S., Kapoor, S., Zschiegner, J., Maddix, D.C., Mahoney, M.W., Torkkola, K., Gor- don Wilson, A., Bohlke-Schneider, M., Wang, Y., 2024. Chronos: Learning the language of time series. Transactions on Ma- chine Learning Research URL:https://openreview.n...

    work page 2024

  4. [4]

    Ex- plainable and interpretable machine learning and data mining

    Atzmueller, M., Fürnkranz, J., Kliegr, T., Schmid, U., 2024. Ex- plainable and interpretable machine learning and data mining. Data Mining and Knowledge Discovery 38, 2571–2595. doi:10.1007/ s10618-024-01041-y

    work page 2024

  5. [5]

    Time series prediction under distri- butionshiftusingdifferentiableforgetting,in:ICML2022Workshop on Principles of Distribution Shift

    Bennett, S., Clarkson, J., 2022. Time series prediction under distri- butionshiftusingdifferentiableforgetting,in:ICML2022Workshop on Principles of Distribution Shift. doi:10.48550/arXiv.2207.11486

    work page doi:10.48550/arxiv.2207.11486 2022

  6. [6]

    Msgnet: learning multi-scale inter-series correlations for multivariate time series forecasting, in: Proceedings of the 38th AAAI Conference on Artificial Intelligence, pp

    Cai, W., Liang, Y., Liu, X., Feng, J., Wu, Y., 2024a. Msgnet: learning multi-scale inter-series correlations for multivariate time series forecasting, in: Proceedings of the 38th AAAI Conference on Artificial Intelligence, pp. 11141–11149. doi:10.1609/aaai.v38i10. 28991

    work page doi:10.1609/aaai.v38i10

  7. [7]

    Cai,Y.,Goswami,M.,Choudhry,A.,Srinivasan,A.,Dubrawski,A.,

    work page

  8. [8]

    URL:https://openreview

    Jolt: Jointly learned representations of language and time- series, in: Proceedings of the 1st Workshop on Deep Generative Models for Health at NeurIPS 20233. URL:https://openreview. net/forum?id=UVF1AMBj9u

    work page

  9. [9]

    Continuous temporal domain generalization, in: Proceedings of the 38th In- ternational Conference on Neural Information Processing Systems

    Cai, Z., Bai, G., Jiang, R., Song, X., Zhao, L., 2024b. Continuous temporal domain generalization, in: Proceedings of the 38th In- ternational Conference on Neural Information Processing Systems. doi:10.5555/3737916.3741980

    work page doi:10.5555/3737916.3741980

  10. [10]

    Science Robotics 8, eadc8892

    Chahine,M.,Hasani,R.,Kao,P.,Ray,A.,Shubert,R.,Lechner,M., Amini,A.,Rus,D.,2023.Robustflightnavigationoutofdistribution with liquid neural networks. Science Robotics 8, eadc8892. doi:10. 1126/scirobotics.adc8892

    work page 2023

  11. [11]

    Align and fine-tune: Enhancing LLMs for time-series forecasting, in: NeurIPS Workshop on Time Series in the Age of Large Models

    Chang, C., Wang, W.Y., Peng, W.C., Chen, T.F., Samtani, S., 2024. Align and fine-tune: Enhancing LLMs for time-series forecasting, in: NeurIPS Workshop on Time Series in the Age of Large Models. URL:https://openreview.net/forum?id=AaRCmJieG4

    work page 2024

  12. [12]

    Chen, M., Shen, L., Fu, H., Li, Z., Sun, J., Liu, C., 2024. Cali- bration of time-series forecasting: Detecting and adapting context- drivendistributionshift,in:Proceedingsofthe30thACMSIGKDD ConferenceonKnowledgeDiscoveryandDataMining,p.341–352. doi:10.1145/3637528.3671926

    work page doi:10.1145/3637528.3671926 2024

  13. [13]

    Sheng, Xiaoqiang Qiao, Athanasios V

    Cohen, L., Avrahami-Bakish, G., Last, M., Kandel, A., Kipersztok, O.,2008. Real-timedataminingofnon-stationarydatastreamsfrom sensor networks. Information Fusion 9, 344–353. doi:10.1016/j. inffus.2005.05.005

    work page doi:10.1016/j 2008

  14. [14]

    Autoaugment: Learning augmentation strategies from data, in: Pro- ceedings of the 32nd IEEE/CVF Conference on Computer Vision andPatternRecognition,pp.113–123

    Cubuk, E.D., Zoph, B., Mané, D., Vasudevan, V., Le, Q.V., 2019. Autoaugment: Learning augmentation strategies from data, in: Pro- ceedings of the 32nd IEEE/CVF Conference on Computer Vision andPatternRecognition,pp.113–123. doi:10.1109/CVPR.2019.00020

    work page doi:10.1109/cvpr.2019.00020 2019

  15. [15]

    Recent emerging techniques in explainable artificial intelligence to enhancetheinterpretableandunderstandingofaimodelsforhuman

    Daniel, E., Deborah, U., Anayo, C., Pamela, E., Ngozi, F., 2025. Recent emerging techniques in explainable artificial intelligence to enhancetheinterpretableandunderstandingofaimodelsforhuman. Neural Processing Letters 57, 16. doi:10.1007/s11063-025-11732-2

    work page doi:10.1007/s11063-025-11732-2 2025

  16. [16]

    A decoder-only foundation model for time-series forecasting, in: Proceedings of the 41st International Conference on Machine Learning

    Das, A., Kong, W., Sen, R., Zhou, Y., 2024. A decoder-only foundation model for time-series forecasting, in: Proceedings of the 41st International Conference on Machine Learning. doi:10.5555/ 3692070.3692474

    work page arXiv 2024

  17. [17]

    A continual learning survey: Defying forgetting in classification tasks

    De Lange, M., Aljundi, R., Masana, M., Parisot, S., Jia, X., Leonardis, A., Slabaugh, G., Tuytelaars, T., 2021. A continual learning survey: Defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 3366–3385. doi:10.1109/TPAMI.2021.3057446. X. Wu et al.:Preprint submitted to ElsevierPage 15 of 21 Out-o...

    work page doi:10.1109/tpami.2021.3057446 2021

  18. [18]

    Deng, A., Li, S., Xiong, M., Chen, Z., Hooi, B., 2022. Trust, but verify: Using self-supervised probing to improve trustworthiness, in: Computer Vision - ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XIII, p. 361–377. doi:10.1007/978-3-031-19778-9_21

    work page doi:10.1007/978-3-031-19778-9_21 2022

  19. [19]

    Disentangling structured components: Towards adap- tive, interpretable and scalable time series forecasting

    Deng, J., Chen, X., Jiang, R., Yin, D., Yang, Y., Song, X., Tsang, I.W., 2024a. Disentangling structured components: Towards adap- tive, interpretable and scalable time series forecasting. IEEE Trans- actionsonKnowledgeandDataEngineering36,3783–3800. doi:10. 1109/TKDE.2024.3371931

    work page arXiv 2024

  20. [20]

    Domaingeneralizationintimeseriesforecasting.ACMTransactions on Knowledge Discovery from Data 18

    Deng, S., Sprangers, O., Li, M., Schelter, S., de Rijke, M., 2024b. Domaingeneralizationintimeseriesforecasting.ACMTransactions on Knowledge Discovery from Data 18. doi:10.1145/3643035

    work page doi:10.1145/3643035

  21. [21]

    URL:http://poster-openaccess.com/files/ICIC2025/ 3586.pdf

    Deng,Z.,Wu,Z.,Su,L.,Wu,Y.,Zheng,Q.,.Alleviatingdistribution shift in time series forecasting with an invertible neural network transformation, in: 2025 International Conference on Intelligent Computing. URL:http://poster-openaccess.com/files/ICIC2025/ 3586.pdf

    work page 2025

  22. [22]

    Explainable ai for clinical and remote health applications: a survey on tabular and time series data

    Di Martino, F., Delmastro, F., 2023. Explainable ai for clinical and remote health applications: a survey on tabular and time series data. Artificial Intelligence Review 56, 5261–5315. doi:10.1007/ s10462-022-10304-3

    work page 2023

  23. [23]

    Rethinking adam for time series forecasting: A simple heuristic to improve optimization under distribution shifts

    Dong, Y., Wu, J., . Rethinking adam for time series forecasting: A simple heuristic to improve optimization under distribution shifts. Available at SSRN 5570331 doi:10.2139/ssrn.5570331

    work page doi:10.2139/ssrn.5570331

  24. [24]

    Dooley, S., Khurana, G.S., Mohapatra, C., Naidu, S., White, C.,

    work page

  25. [25]

    doi:10.5555/3666122.3666234

    Forecastpfn: synthetically-trained zero-shot forecasting, in: Proceedings of the 37th International Conference on Neural Infor- mation Processing Systems. doi:10.5555/3666122.3666234

    work page doi:10.5555/3666122.3666234

  26. [26]

    Rethinkingvideoanomalydetection-a continuallearningapproach,in:Proceedingsofthe22ndIEEE/CVF Winter Conference on Applications of Computer Vision, pp

    Doshi,K.,Yilmaz,Y.,2022. Rethinkingvideoanomalydetection-a continuallearningapproach,in:Proceedingsofthe22ndIEEE/CVF Winter Conference on Applications of Computer Vision, pp. 3036–

    work page 2022

  27. [27]

    doi:10.1109/WACV51458.2022.00309

    work page doi:10.1109/wacv51458.2022.00309 2022

  28. [28]

    Deep air quality fore- casting using hybrid deep learning framework

    Du, S., Li, T., Yang, Y., Horng, S.J., 2021. Deep air quality fore- casting using hybrid deep learning framework. IEEE Transactions on Knowledge and Data Engineering 33, 2412–2424. doi:10.1109/ TKDE.2019.2954510

    work page arXiv 2021

  29. [29]

    electric vehicles

    Du,S.,Zhang,J.,Wang,Y.,Li,Z.,2024. Integrationofcomputervi- sionandiotintoanautomaticdrivingassistancesystemfor“electric vehicles”. IEEE Transactions on Industrial Informatics 20, 4765–

    work page 2024

  30. [30]

    doi:10.1109/TII.2023.3326546

    work page doi:10.1109/tii.2023.3326546 2023

  31. [31]

    Combating distribution shift for accurate time series forecasting via hypernet- works,in:2022IEEE28thInternationalConferenceonParalleland Distributed System, pp

    Duan, W., He, X., Zhou, L., Thiele, L., Rao, H., 2023. Combating distribution shift for accurate time series forecasting via hypernet- works,in:2022IEEE28thInternationalConferenceonParalleland Distributed System, pp. 900–907. doi:10.1109/ICPADS56603.2022. 00121

    work page doi:10.1109/icpads56603.2022 2023

  32. [32]

    IEEETransactionsonKnowledgeandDataEngineering 35, 10339–10350

    Dudek,G.,2023.Std:Aseasonal-trend-dispersiondecompositionof timeseries. IEEETransactionsonKnowledgeandDataEngineering 35, 10339–10350. doi:10.1109/TKDE.2023.3268125

    work page doi:10.1109/tkde.2023.3268125 2023

  33. [33]

    Ekambaram, V., Jati, A., Dayama, P., Mukherjee, S., Nguyen, N.H., Gifford, W.M., Reddy, C., Kalagnanam, J., 2024. Tiny time mixers (ttms): fast pre-trained models for enhanced zero/few-shot forecast- ing of multivariate time series, in: Proceedings of the 38th Inter- national Conference on Neural Information Processing Systems. doi:10.5555/3737916.3740275

    work page doi:10.5555/3737916.3740275 2024

  34. [34]

    Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting, in: Proceedings of the 29th ACM SIGKDD ConferenceonKnowledgeDiscoveryandDataMining,p.459–469

    Ekambaram, V., Jati, A., Nguyen, N., Sinthong, P., Kalagnanam, J., 2023. Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting, in: Proceedings of the 29th ACM SIGKDD ConferenceonKnowledgeDiscoveryandDataMining,p.459–469. doi:10.1145/3580305.3599533

    work page doi:10.1145/3580305.3599533 2023

  35. [35]

    Medvia: Empowering medical time series classification with vision augmentation and multimodal fusion

    Fan,W.,Fei,J.,Han,J.,Lian,J.,Ye,H.,Song,X.,Lv,X.,Yi,K.,Li, M., 2025a. Medvia: Empowering medical time series classification with vision augmentation and multimodal fusion. Information Fusion , 103659doi:10.1016/j.inffus.2025.103659

    work page doi:10.1016/j.inffus.2025.103659 2025

  36. [36]

    Dish-ts: a general paradigm for alleviating distribution shift in time series forecasting, in: Proceedings of the 37th AAAI Conference on Artificial Intelligence

    Fan, W., Wang, P., Wang, D., Wang, D., Zhou, Y., Fu, Y., 2023. Dish-ts: a general paradigm for alleviating distribution shift in time series forecasting, in: Proceedings of the 37th AAAI Conference on Artificial Intelligence. doi:10.1609/aaai.v37i6.25914

    work page doi:10.1609/aaai.v37i6.25914 2023

  37. [37]

    In-flow: Instance normalization flow for non- stationary time series forecasting, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.1, p

    Fan, W., Zheng, S., Wang, P., Xie, R., Yi, K., Zhang, Q., Bian, J., Fu, Y., 2025b. In-flow: Instance normalization flow for non- stationary time series forecasting, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.1, p. 295–306. doi:10.1145/3690624.3709260

    work page doi:10.1145/3690624.3709260

  38. [38]

    Feng, C., Huang, L., Krompass, D., 2024a. General time trans- former:anencoder-onlyfoundationmodelforzero-shotmultivariate time series forecasting, in: Proceedings of the 33rd ACM Interna- tional Conference on Information and Knowledge Management, p. 3757–3761. doi:10.1145/3627673.3679931

    work page doi:10.1145/3627673.3679931

  39. [39]

    A causal representation learning based model for time series prediction under external interference

    Feng, X., Fan, D., Jiang, S., Zhang, J., Guo, B., Ding, X., Hu, D., Jiang, Y., 2024b. A causal representation learning based model for time series prediction under external interference. Information Sciences 663, 120270. doi:10.1016/j.ins.2024.120270

    work page doi:10.1016/j.ins.2024.120270 2024

  40. [40]

    A new approach to interoperability withinthesmartcitybasedontimeseries-embeddedadaptivetraffic prediction modelling

    Fernandez, V., Pérez, V., 2024. A new approach to interoperability withinthesmartcitybasedontimeseries-embeddedadaptivetraffic prediction modelling. Networks and Spatial Economics -, 1–16. doi:10.1007/s11067-024-09662-y

    work page doi:10.1007/s11067-024-09662-y 2024

  41. [41]

    Information fusion in wirelesssensornetworkswithsourcecorrelation.Informationfusion 15, 80–89

    Ferrari, G., Martalò, M., Abrardo, A., 2014. Information fusion in wirelesssensornetworkswithsourcecorrelation.Informationfusion 15, 80–89. doi:10.1016/j.inffus.2012.09.001

    work page doi:10.1016/j.inffus.2012.09.001 2014

  42. [42]

    Deep learning with long short-term memorynetworksforfinancialmarketpredictions.EuropeanJournal ofOperationalResearch270,654–669

    Fischer, T., Krauss, C., 2018. Deep learning with long short-term memorynetworksforfinancialmarketpredictions.EuropeanJournal ofOperationalResearch270,654–669. doi:10.1016/j.ejor.2017.11. 054

    work page doi:10.1016/j.ejor.2017.11 2018

  43. [43]

    Catsight, a direct path to proper multi-variate time series changedetection:perceivingaconceptdriftthroughcommonspatial pattern

    Florez,A.,Rodriguez-Moreno,I.,Artetxe,A.,Olaizola,I.G.,Sierra, B., 2023. Catsight, a direct path to proper multi-variate time series changedetection:perceivingaconceptdriftthroughcommonspatial pattern. InternationalJournalofMachineLearningandCybernetics 14, 2925–2944. doi:10.1007/s13042-023-01810-z

    work page doi:10.1007/s13042-023-01810-z 2023

  44. [44]

    Tackling concept drift by temporal inductive transfer, in: Proceedings of the 29th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, p

    Forman, G., 2006. Tackling concept drift by temporal inductive transfer, in: Proceedings of the 29th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, p. 252–259. doi:10.1145/1148170.1148216

    work page doi:10.1145/1148170.1148216 2006

  45. [45]

    Remotesensingtimeseriesanalysis:A reviewofdataandapplications

    Fu, Y., Zhu, Z., Liu, L., Zhan, W., He, T., Shen, H., Zhao, J., Liu, Y.,Zhang,H.,Liu,Z.,2024. Remotesensingtimeseriesanalysis:A reviewofdataandapplications. JournalofRemoteSensing4,0285. doi:10.34133/remotesensing.0285

    work page doi:10.34133/remotesensing.0285 2024

  46. [46]

    A survey on concept drift adaptation

    Gama, J.a., Žliobaitundefined, I., Bifet, A., Pechenizkiy, M., Bouchachia, A., 2014. A survey on concept drift adaptation. ACM Computing Surveys 46. doi:10.1145/2523813

    work page doi:10.1145/2523813 2014

  47. [47]

    Holzinger, G

    Gan, W., Lin, J.C., Chao, H., Zhan, J., 2017. Data mining in distributedenvironment:asurvey. WileyInterdisciplinaryReviews: DataMiningandKnowledgeDiscovery7,e1216. doi:10.1002/widm. 1216

    work page doi:10.1002/widm 2017

  48. [48]

    Gao, H., Dong, Z., Yong, J., Fukushima, S., Taura, K., Jiang, R.,

    work page

  49. [49]

    doi:10.48550/arXiv.2510.04908

    How different from the past? spatio-temporal time series forecasting with self-supervised deviation learning, in: Proceedings of the 39th Conference on Neural Information Processing Systems. doi:10.48550/arXiv.2510.04908

    work page doi:10.48550/arxiv.2510.04908

  50. [50]

    UniTS: A unified multi-task time series model, in: Proceedings of the 38th Annual Conference on Neural Informa- tion Processing Systems

    Gao, S., Koker, T., Queen, O., Hartvigsen, T., Tsiligkaridis, T., Zitnik, M., 2024. UniTS: A unified multi-task time series model, in: Proceedings of the 38th Annual Conference on Neural Informa- tion Processing Systems. URL:https://openreview.net/forum?id= nBOdYBptWW

    work page 2024

  51. [51]

    Garcia,C.M.,Abilio,R.,Koerich,A.L.,Britto,A.d.S.,Barddal,J.P.,

    work page

  52. [52]

    ACM Transactions on Intelligent Systems and Technology 16

    Concept drift adaptation in text stream mining settings: A systematic review. ACM Transactions on Intelligent Systems and Technology 16. doi:10.1145/3704922

    work page doi:10.1145/3704922

  53. [53]

    Timegpt-1,

    Garza,A.,Challu,C.,Mergenthaler-Canseco,M.,2024. Timegpt-1. doi:10.48550/arXiv.2310.03589,arXiv:2310.03589

    work page doi:10.48550/arxiv.2310.03589 2024

  54. [54]

    Conformal inference for online pre- diction with arbitrary distribution shifts

    Gibbs, I., Candès, E., 2024. Conformal inference for online pre- diction with arbitrary distribution shifts. The Journal of Machine Learning Research 25. doi:10.5555/3722577.3722739

    work page doi:10.5555/3722577.3722739 2024

  55. [55]

    Conformal inference for online prediction with arbitrary distribution shifts

    Gibbs, I., Candès, E.J., 2024. Conformal inference for online prediction with arbitrary distribution shifts. Journal of Machine Learning Research 25, 1–36. URL:http://jmlr.org/papers/v25/ 22-1218.html. X. Wu et al.:Preprint submitted to ElsevierPage 16 of 21 Out-of-Distribution Generalization in Time Series: A Survey

    work page 2024

  56. [57]

    Göring, N., Hess, F., Brenner, M., Monfared, Z., Durstewitz, D.,

    work page

  57. [58]

    doi:10.5555/3692070.3692711

    Out-of-domain generalization in dynamical systems recon- struction, in: Proceedings of the 41st International Conference on Machine Learning. doi:10.5555/3692070.3692711

    work page doi:10.5555/3692070.3692711

  58. [59]

    doi:10.5555/3692070.3692711

    Goswami, M., Szafer, K., Choudhry, A., Cai, Y., Li, S., Dubrawski, A.,2024. Moment:afamilyofopentime-seriesfoundationmodels, in: Proceedings of the 41st International Conference on Machine Learning. doi:10.5555/3692070.3692712

    work page doi:10.5555/3692070.3692712 2024

  59. [60]

    Large language models are zero-shot time series forecasters, in: Proceedings of the 37th International Conference on Neural Information Processing Systems

    Gruver, N., Finzi, M., Qiu, S., Wilson, A.G., 2023. Large language models are zero-shot time series forecasters, in: Proceedings of the 37th International Conference on Neural Information Processing Systems. doi:10.5555/3666122.3666983

    work page doi:10.5555/3666122.3666983 2023

  60. [61]

    8029–8037

    Gunasekara, N., Pfahringer, B., Murilo Gomes, H., Bifet, A., Koh, Y.S.,2024.Recurrentconceptdriftsondatastreams,in:Proceedings ofthe33rdInternationalJointConferenceonArtificialIntelligence, pp. 8029–8037. doi:10.24963/ijcai.2024/888

    work page doi:10.24963/ijcai.2024/888 2024

  61. [62]

    H. F. M. Oliveira, G., C. Cavalcante, R., G. Cabral, G., L. Minku, L., L. I. Oliveira, A., 2017. Time series forecasting in the presence ofconceptdrift:Apso-basedapproach,in:2017IEEE29thInterna- tionalConferenceonToolswithArtificialIntelligence,pp.239–246. doi:10.1109/ICTAI.2017.00046

    work page doi:10.1109/ictai.2017.00046 2017

  62. [63]

    Artificial intelligence techniques for driving safety and vehicle crash predic- tion

    Halim, Z., Kalsoom, R., Bashir, S., Abbas, G., 2016. Artificial intelligence techniques for driving safety and vehicle crash predic- tion. Artificial Intelligence Review 46, 351–387. doi:10.1007/ s10462-016-9467-9

    work page 2016

  63. [64]

    Out-of-distribution detection-assisted trust- worthy machinery fault diagnosis approach with uncertainty-aware deep ensembles

    Han, T., Li, Y.F., 2022. Out-of-distribution detection-assisted trust- worthy machinery fault diagnosis approach with uncertainty-aware deep ensembles. Reliability Engineering and System Safety 226, 108648. doi:10.1016/j.ress.2022.108648

    work page doi:10.1016/j.ress.2022.108648 2022

  64. [65]

    Domain adaptation for time series under feature and labelshifts,in:Proceedingsofthe40thInternationalConferenceon Machine Learning

    He, H., Queen, O., Koker, T., Cuevas, C., Tsiligkaridis, T., Zitnik, M., 2023. Domain adaptation for time series under feature and labelshifts,in:Proceedingsofthe40thInternationalConferenceon Machine Learning. doi:10.5555/3618408.3618926

    work page doi:10.5555/3618408.3618926 2023

  65. [66]

    Dis- tributional drift adaptation with temporal conditional variational autoencoder for multivariate time series forecasting

    He, H., Zhang, Q., Yi, K., Shi, K., Niu, Z., Cao, L., 2025a. Dis- tributional drift adaptation with temporal conditional variational autoencoder for multivariate time series forecasting. IEEE Trans- actions on Neural Networks and Learning Systems 36, 7287–7301. doi:10.1109/TNNLS.2024.3384842

    work page doi:10.1109/tnnls.2024.3384842 2024

  66. [67]

    Robust multivariate time series forecasting against intraseries and interseriestransitionalshift

    He,H.,Zhang,Q.,Yi,K.,Xue,X.,Wang,S.,Hu,L.,Cao,L.,2025b. Robust multivariate time series forecasting against intraseries and interseriestransitionalshift. IEEETransactionsonNeuralNetworks and Learning Systems -, 1–14. doi:10.1109/TNNLS.2025.3593156

    work page doi:10.1109/tnnls.2025.3593156 2025

  67. [68]

    Hou, S., Qin, Z., Wang, J., An, J., Zhang, Y., Wang, X., Wang, Z.,

    work page

  68. [69]

    Multi-granularity feature fusion network for cross-domain person re-identification, in: Proceedings of the 14th International Conference on Information Science and Technology, pp. 840–846. doi:10.1109/ICIST63249.2024.10805357

    work page doi:10.1109/icist63249.2024.10805357 2024

  69. [70]

    Hu, X., Fan, W., Yi, K., Wang, P., Xu, Y., Fu, Y., Wang, P.,

    work page

  70. [71]

    Boosting urban prediction via addressing spatial-temporal distribution shift, in: 2023 IEEE International Conference on Data Mining, pp. 160–169. doi:10.1109/ICDM58522.2023.00025

    work page doi:10.1109/icdm58522.2023.00025 2023

  71. [72]

    Varadarajan, A

    Hu, Y., Jia, X., Tomizuka, M., Zhan, W., 2022. Causal-based time series domain generalization for vehicle intention prediction, in: Proceedings of the 39th International Conference on Robotics and Automation, pp. 7806–7813. doi:10.1109/ICRA46639.2022.9812188

    work page doi:10.1109/icra46639.2022.9812188 2022

  72. [73]

    Huang, S., Zhang, T., Zhang, Z., Wang, X., Wang, L., Wang, X.,

    work page

  73. [74]

    991–1002

    Metaeformer: Unveiling and leveraging meta-patterns for complex and dynamic systems load forecasting, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2, p. 991–1002. doi:10.1145/3711896.3737047

    work page doi:10.1145/3711896.3737047

  74. [75]

    Time-llm: Time series forecasting by reprogramming large language models, in: Proceedings of the 12th International Conference on Learning Rep- resentations

    Jin, M., Wang, S., Ma, L., Chu, Z., Zhang, J.Y., Shi, X., Chen, P.Y., Liang, Y., Li, Y.F., Pan, S., Wen, Q., 2024a. Time-llm: Time series forecasting by reprogramming large language models, in: Proceedings of the 12th International Conference on Learning Rep- resentations. URL:https://openreview.net/forum?id=Unb5CVPtae

    work page

  75. [76]

    Position: what can large language models tell us about time series analysis, in: Proceedings of the 41st International Conference on Machine Learning

    Jin,M.,Zhang,Y.,Chen,W.,Zhang,K.,Liang,Y.,Yang,B.,Wang, J., Pan, S., Wen, Q., 2024b. Position: what can large language models tell us about time series analysis, in: Proceedings of the 41st International Conference on Machine Learning. doi:10.5555/ 3692070.3692965,

    work page arXiv

  76. [77]

    Temporaldomaingeneral- izationvialearninginstance-levelevolvingpatterns,in:Proceedings ofthe33rdInternationalJointConferenceonArtificialIntelligence

    Jin,Y.,Yang,Z.,Chu,X.,Ma,L.,2024c. Temporaldomaingeneral- izationvialearninginstance-levelevolvingpatterns,in:Proceedings ofthe33rdInternationalJointConferenceonArtificialIntelligence. doi:10.24963/ijcai.2024/470

    work page doi:10.24963/ijcai.2024/470 2024

  77. [78]

    K, S.M., S, R.M., S, A.J., R, V.K., 2025. A novel concept drift detection model for handling evolving patterns in multivariate time series, in: 2025 International Conference on Advancements in Power, Communication and Intelligent Systems, pp. 1–6. doi:10. 1109/APCI65531.2025.11136854

    work page arXiv 2025

  78. [79]

    Artificial Intelligence Review 57, 285

    Khoee,A.G.,Yu,Y.,Feldt,R.,2024.Domaingeneralizationthrough meta-learning: a survey. Artificial Intelligence Review 57, 285. doi:10.1007/s10462-024-10922-z

    work page doi:10.1007/s10462-024-10922-z 2024

  79. [80]

    Learnability and algorithm for continual learning, in: Proceedings of the 40th Inter- national Conference on Machine Learning

    Kim, G., Xiao, C., Konishi, T., Liu, B., 2023. Learnability and algorithm for continual learning, in: Proceedings of the 40th Inter- national Conference on Machine Learning. doi:10.5555/3618408. 36191023

    work page doi:10.5555/3618408 2023

  80. [81]

    Kim, T., Kim, J., Tae, Y., Park, C., Choi, J.H., Choo, J., 2022. Reversible instance normalization for accurate time-series forecast- ing against distribution shift, in: Proceedings of the 10th Interna- tional Conference on Learning Representations. URL:https:// openreview.net/forum?id=cGDAkQo1C0p

    work page 2022

Showing first 80 references.