Pith. sign in

REVIEW 4 major objections 5 minor 80 references

A Wave is Worth 100 Words: Investigating Cross-Domain Transferability in Time Series

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims that time series from many domains can be mapped into one shared spectral space by wavelet quantization, so a single pretrained transformer handles forecasting, imputation, and classification across domains with few or…

desk verdict The empirical sweep is real, but the load-bearing bijection theorem is proved for an infinite basis, not the finite wavebook the method actually uses. read the letter →

arxiv 2412.00772 v1 pith:FPARZV6A submitted 2024-12-01 cs.LG cs.AI

Xiangkai Ma , Xiaobin Hong , Wenzhong Li , Sanglu Lu This is my paper · ORCID
classification cs.LGcs.AI MSC 68T0742C40
keywords timeseriescross-domaintransferwavequantizationtokenizationspectrallatentspacepretrainingzero-shotlearningWQ4TS
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

The paper argues that the main obstacle to cross-domain time series learning is that raw series from different domains look nothing alike, so a model trained on one source fails on another. It proposes WQ4TS, a pretraining method that projects every series into a common spectral latent space built from a fixed set of wavelet basis functions, the 'wavebook', turning each local window of the series into a token of fluctuation-pattern similarities. A single Transformer encoder pretrained in that shared space is then applied to forecasting, imputation, and classification across datasets with different sampling rates, periodicities, and scales, under full-data, few-shot, or zero-shot supervision. If the claim holds, one pretrained model could replace per-dataset training, which matters because labeled time series data is costly while unlabeled series are abundant.

What carries the argument

The wavebook: a set of $\lambda$ orthogonal wavelet basis functions $\{A_1, \ldots, A_{\lambda}\}$ obtained by rescaling one orthogonal wavelet at scale factors $S_i = 2 f_c \lambda / i$, together spanning a $\lambda$-dimensional common spectral latent space $V^{\lambda}$. The tokenization computes, for each timestep $j$, the fluctuation pattern similarity between the local window around $j$ and each basis function, producing a token in $\mathbb{R}^{\lambda}$; the tokens are the coordinates of the series in $V^{\lambda}$. This machinery does the work of the argument: it is what claims to make cross-domain transfer possible, since all domains are projected through the same fixed basis set, and it is what the paper claims to be interpretable and bijective, with each token carrying semantic pattern information and sub-series and coordinates corresponding one-to-one.

What would settle it

Run the tokenization of Equations (13)--(14) on all test windows of two datasets with very different dynamics, such as ETTh1 and Weather, and attempt to invert the finite $\lambda$-dimensional projection, or simply count how often distinct segments produce near-identical token vectors. If the reconstruction error is substantial, or token collisions are frequent as $\lambda$ shrinks, the claimed bijection and the semantic-content guarantee would be refuted.

Watch

Extended reading notes

Core claim

WQ4TS introduces the 'wave as token' principle: a set of orthogonal wavelet basis functions (the wavebook) forms a $\lambda$-dimensional common space $V^{\lambda}$, and each local segment of an input series is mapped to coordinates in that space by an inner-product convolution, so $\text{token}_j = (p_{1,j}, \ldots, p_{\lambda,j})$ records the pattern similarity between the segment at timestep $j$ and each basis function. Because every domain is projected through the same fixed wavebook, the paper claims the resulting tokens live in one shared, interpretable space with semantic content, and that the sub-series-to-coordinates map is bijective. A vanilla Transformer encoder is pretrained on tokens from multiple domains with a learnable multi-task weighting, then full-parameter fine-tuned on the target domain. In experiments across seven forecasting/imputation datasets and 35 UCR classification datasets, WQ4TS reports the best performance on 87.5% of all tasks, average metric improvement up to 34.7%, and 25.8% and 44.1% improvements in few-shot and zero-shot settings; multi-domain pretraining beats single-domain pretraining, which the paper reads as evidence that the common space mitigates negative migration.

Load-bearing premise

The whole method assumes that projecting a series onto a fixed finite set of wavelet shapes (the wavebook) preserves each local pattern without loss, so that tokens carry the same information in every domain, but the paper proves lossless bijection only for an infinite wavelet family, not for the finite wavebook actually used.

Editorial extensions

If this is right

  • A single pretrained model can be deployed on a new time series dataset with no retraining on that domain's data, bridging differences in sampling rate, channel count, and periodicity.
  • Few-shot performance with 5% of target training data approaches what current specialized models achieve with full data, lowering the annotation cost for forecasting, imputation, and classification.
  • Adding the wave quantize module improves existing models such as DLinear, PatchTST, FEDformer, Autoformer, OneFitsAll, and TimesNet, so the transfer benefit is not tied to the WQ4TS architecture itself.
  • Multi-domain pretraining outperforms single-domain pretraining, indicating that the common space reduces negative migration rather than injecting noise from unrelated domains.

Reading between the lines

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

  • Inference: if the tokens are genuinely shared coordinates, the same framework should extend to anomaly detection and regression on series with very different physical meanings, such as ECG to power load, a setting the paper does not test.
  • Inference: the bijection claim, if it held, would make the token stream a lossless encoding and suggest generative decoding from tokens is possible; the paper never reconstructs series from tokens, so reconstruction error is a natural untested check.
  • Inference: because the finite wavebook is a projection, token collisions should increase as $\lambda$ shrinks; measuring collision rates across the paper's datasets would quantify how much semantic information is actually preserved.
  • Inference: the analogy to subword tokenization implies attention operates on roughly equal-information units, so token-importance analysis could identify which waveform patterns drive cross-domain transfer, an interpretability study the paper does not perform.
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 / 5 minor

Summary. The paper proposes WQ4TS, a cross-domain time-series pretraining method built on a wavelet-based tokenization. A 'wavebook' of λ=100 sampled, scaled wavelet filters is used to map each local segment of a series to a λ-dimensional token of 'pattern similarities'; the tokens are fed to an encoder-only Transformer that is pretrained across domains and then fine-tuned or applied zero-shot to forecasting, imputation, and classification. The paper claims that the tokenization is bijective, that tokens carry semantic information in a common spectral latent space, and that WQ4TS achieves the best performance on 87.5% of tasks with an average improvement up to 34.7%.

Significance. The empirical scope is broad and the central idea—a generic, model-agnostic wavelet tokenization that reduces cross-domain heterogeneity—is interesting. The paper evaluates three tasks, three data-scarcity settings, multiple forecasting benchmarks, and 35 UCR classification datasets, and it includes an ablation and a plug-in study over existing backbones. However, the theoretical foundation for the common spectral space is not established for the implemented finite wavebook, and the reproducibility of the experiments is limited by missing implementation details, absent error bars, and no released code. If the theoretical claims were corrected or substantially weakened, the empirical study could still be a useful contribution; as written, the load-bearing justification for the method's cross-domain transfer mechanism is unsupported.

major comments (4)
  1. [Sections 3.2–3.3, Proposition 2 and Eqs. (11)–(14)] Proposition 2 proves coefficient uniqueness only for an infinite orthonormal wavelet basis of L2(R). The implemented map in Eqs. (11)–(14) uses λ=100 filters A_i of different lengths that are resampled from a single amplitude sequence; this is a finite linear map from a local segment to R^λ, with no proof that the segment dimension equals λ or that the filter matrix is invertible. In fact, filters of different lengths and scales are generically not mutually orthogonal. The claims that 'the bijection relation is satisfied between the sub-series and the coordinates' and that each token has 'semantic information' therefore do not follow. This is load-bearing because the common spectral latent space and the cross-domain transfer story rest on this bijection.
  2. [Section 3.2 and Algorithm 1] The wavebook construction is underspecified. The manuscript never states which orthogonal wavelet F(t) is used, which amplitude sequence A∈R^{2m} is chosen, what precision m is, what central frequency fc is, or how F(t) is 'designed to adapt to the characteristic of the data domain' as Algorithm 1 instructs. Since these choices completely determine the basis functions in Eqs. (11)–(12), the method is not reproducible as described, and the sensitivity of the results to these choices is unknown.
  3. [Section 5.1 and Tables 4–10] Although Section 5.1 states that all experiments are repeated three times, no error bars, standard deviations, or per-seed results are reported anywhere, and no code is released. This matters particularly because several reported gaps are very large, e.g., zero-shot imputation ETTm1→Exchange in Table 9 reports MSE 0.003 for WQ4TS versus 3.107 for FEDformer. Without uncertainty estimates or code, the robustness and fairness of these comparisons cannot be verified.
  4. [Section 3.3, Eq. (14)] The definition p_{i,j}=d_{i,j+n/2-1} requires n=m·S_i to be even, but no evenness constraint is stated for the scale factors S_i. For odd n, the index j+n/2-1 is not an integer and the expression is undefined. In addition, the effective 'local segment' length is never defined, and it varies with i because each filter A_i has a different length; this further undermines the interpretation of token_j as a coordinate of a single fixed sub-series.
minor comments (5)
  1. [Table 4 vs Table 16] The baseline called OneFitsAll in Table 4 appears as GPT4TS in Table 16 with identical numbers; the naming should be consistent.
  2. [Table 1] The construction of the 'Resample dataset' is not described; the caption mentions rate=10 but the resampling procedure and the relation to the original sampling rate are unclear.
  3. [Section 3, introduction] The text says the tokenization has 'no specific requirements for hyperparameters', yet λ, m, fc, and the number of encoder layers L are free hyperparameters listed in Table 3; the claim should be qualified as 'no dataset-specific tuning'.
  4. [Proof of Proposition 2, Eq. (1)] The notation |F_{m,n}(t)|^2 in the denominator should be the squared L2 norm of F_{m,n}, not a pointwise absolute value.
  5. [Abstract and Table 16] The abstract states 'best performance on 87.5% of all tasks', but Table 16's '1st Count' shows 78 wins out of 80 metric entries (97.5%); the discrepancy between 'tasks' and 'metric entries' should be clarified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical comparisons are external-benchmark based and the theoretical bijection claim, though unsupported for the finite implementation, is not derived by self-citation or by fitting.

full rationale

I find no circular step in the paper's derivation chain. The empirical claims are benchmarked against external baselines (Tables 4-10, 16-27) under full-data, few-shot, and zero-shot settings; the wave quantize tokenizer is a fixed wavelet-projection module with no parameters fitted to test labels or to the reported metrics, so the SOTA comparisons are not forced by construction. The theoretical bijectivity claim (Proposition 2, Sections 3.1-3.3) is supported by the standard infinite-dimensional orthogonal wavelet basis theorem, not by a self-citation: the reference list contains no works by the present authors, so there is no load-bearing self-citation chain and no imported uniqueness theorem from the authors' prior work. The finite implementation in Eqs. (11)-(14) uses only lambda scaled, sampled basis filters and therefore implements a projection rather than the infinite-basis bijection proven in Proposition 2; this is a genuine correctness gap in the paper's stated interpretability and semantic-information claims, but it is not circularity, because the claim is not equivalent to its own input by definition and the empirical results are independently checkable against external baselines. The only internal-consistency flavor is that the semantic-information claim is argued from the same inner-product construction used to define tokens; that is a justification of a property of the construction, not a reduction of a prediction to a fitted parameter. Accordingly, the honest finding is no significant circularity.

Assumptions & free parameters 4 free parameters · 3 assumptions · 1 invented entities

The central claim rests on standard wavelet theory plus an unjustified finite-basis bijection assumption. All other elements are conventional ML hyperparameters or architecture choices, not fitted physical or mathematical constants.

free parameters (4)
  • wavebook size λ = 100 (Table 3)
    The dimension of the shared token space; chosen by hand, not fitted, but central to the method.
  • amplitude sequence precision m = not specified
    Controls the discretization of the wavelet function; no value or selection procedure is given in the paper.
  • central frequency fc = not specified
    Used in Eq. (11) to compute the scale factors; no value or selection procedure is provided.
  • number of encoder layers L = 5-10
    Architecture hyperparameter chosen per task; not fitted to data.
assumptions (3)
  • standard math Standard orthogonal wavelet decomposition of L2(R) (multiresolution analysis and the unitary matrix condition for M(ω))
    Propositions 2 and 3 rely on wavelet basis theory, Shannon sampling, and Mallat's construction; these are standard results.
  • domain assumption The chosen orthogonal wavelet F(t) can be discretized into a finite amplitude sequence A and resampled to form the wavebook
    Not all wavelets admit finite, stable discrete approximations; the paper never specifies F(t) or the discretization error.
  • ad hoc to paper The finite λ-dimensional wavebook yields a bijective representation of any sub-series
    This is the unjustified leap from the infinite wavelet basis to the finite wavebook; no proof is given for the finite case, and it is generally false for arbitrary series.
invented entities (1)
  • Wavebook
    purpose: A fixed set of λ finite-length wavelet basis functions used to tokenize all input series into a common spectral latent space
    The wavebook is an algorithmic construct; no independent evidence is provided that a single finite wavebook preserves information across all domains or that it is interpretable in the claimed sense.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Wave is Worth 100 Words: Investigating Cross-Domain Transferability in Time Series." pith.science (2026). https://pith.science/paper/FPARZV6A

@misc{pith2026241200772,
  author       = {Pith},
  title        = {Pith review of: A Wave is Worth 100 Words: Investigating Cross-Domain Transferability in Time Series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FPARZV6A}},
  note         = {Machine review of arXiv:2412.00772}
}
read the original abstract

Time series analysis is a fundamental data mining task that supervised training methods based on empirical risk minimization have proven their effectiveness on specific tasks and datasets. However, the acquisition of well-annotated data is costly and a large amount of unlabeled series data is under-utilized. Due to distributional shifts across various domains and different patterns of interest across multiple tasks. The problem of cross-domain multi-task migration of time series remains a significant challenge. To address these problems, this paper proposes a novel cross-domain pretraining method based on Wave Quantization (termed as WQ4TS), which can be combined with any advanced time series model and applied to multiple downstream tasks. Specifically, we transfer the time series data from different domains into a common spectral latent space, and enable the model to learn the temporal pattern knowledge of different domains directly from the common space and utilize it for the inference of downstream tasks, thereby mitigating the challenge of heterogeneous cross-domains migration. The establishment of spectral latent space brings at least three benefits, cross-domain migration capability thus adapting to zero- and few-shot scenarios without relying on priori knowledge of the dataset, general compatible cross-domain migration framework without changing the existing model structure, and robust modeling capability thus achieving SOTA results in multiple downstream tasks. To demonstrate the effectiveness of the proposed approach, we conduct extensive experiments including three important tasks: forecasting, imputation, and classification. And three common real-world data scenarios are simulated: full-data, few-shot, and zero-shot. The proposed WQ4TS achieves the best performance on 87.5% of all tasks, and the average improvement of the metrics on all the tasks is up to 34.7%.

Figures

Figures reproduced from arXiv: 2412.00772 by the authors.

Figure 1
Figure 1. Illustration of the proposed WQ4TS architecture. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Model comparison in classification. The results are averaged from 35 subsets of UCR. The proposed WQ4TS [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

80 extracted references · 64 canonical work pages

  1. [1]

    Bousmalis, N

    K. Bousmalis, N. Silberman, D. Dohan, D. Erhan, and D. Krishnan. Unsupervised pixel-level domain adaptation with generative adversarial networks. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 95--104, 2017

    work page 2017

  2. [2]

    T. B. Brown, B. Mann, N. Ryder, and M. S. et al. Language models are few-shot learners. ArXiv, abs/2005.14165, 2020

    arXiv 2005

  3. [3]

    R. Cai, J. Chen, Z. Li, W. Chen, K. Zhang, J. Ye, Z. Li, X. Yang, and Z. Zhang. Time series domain adaptation via sparse associative structure alignment. In AAAI Conference on Artificial Intelligence, 2021

    work page 2021

  4. [4]

    D. Cao, F. Jia, S. \"O . Arik, T. Pfister, Y. Zheng, W. Ye, and Y. Liu. Tempo: Prompt-based generative pre-trained transformer for time series forecasting. ArXiv, abs/2310.04948, 2023

    arXiv 2023

  5. [5]

    Chang, W

    C. Chang, W. Peng, and T.-F. Chen. Llm4ts: Aligning pre-trained llms as data-efficient time-series forecasters. In ArXiv, 2023

    work page 2023

  6. [6]

    C. Chen, Z. Fu, Z. Chen, S. Jin, Z. Cheng, X. Jin, and X. Hua. Homm: Higher-order moment matching for unsupervised domain adaptation. In AAAI Conference on Artificial Intelligence, 2019

    work page 2019

  7. [7]

    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, G. Batista, and Hexagon-ML. The ucr time series classification archive, October 2018. https://www.cs.ucr.edu/ eamonn/time_series_data_2018/

    work page 2018

  8. [8]

    Z. Du, J. Li, H. Su, L. Zhu, and K. Lu. Cross-domain gradient discrepancy minimization for unsupervised domain adaptation. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3936--3945, 2021

    work page 2021

Show all 80 references
  1. [9]

    Y. Fang, Y. Qin, H. Luo, F. Zhao, B. Xu, L. Zeng, and C. Wang. When spatio-temporal meet wavelets: Disentangled traffic forecasting via efficient spectral graph attention networks. ICDE, pages 517--529, 2023

    2023

  2. [10]

    S. E. Finder, Y. Zohav, M. Ashkenazi, and E. Treister. Wavelet feature maps compression for image-to-image cnns. ArXiv, abs/2205.12268, 2022

    2022 arXiv

  3. [11]

    Godfried, K

    I. Godfried, K. Mahajan, M. Wang, K. Li, and P. Tiwari. Flowdb a large scale precipitation, river, and flash flood dataset, 2020

    2020

  4. [12]

    S. Guo, Y. Lin, L. Gong, C. Wang, Z. Zhou, Z. Shen, Y. Huang, and H. Wan. Self-supervised spatial-temporal bottleneck attentive network for efficient long-term traffic forecasting. ICDE, pages 1585--1596, 2023

    2023

  5. [13]

    H. He, O. Queen, T. Koker, C. Cuevas, T. Tsiligkaridis, and M. Zitnik. Domain adaptation for time series under feature and label shifts. In International Conference on Machine Learning, ICML, 2023

    2023

  6. [14]

    Y. Hong, L. Niu, J. Zhang, and L. Zhang. Matchinggan: Matching-based few-shot image generation. In 2020 IEEE International Conference on Multimedia and Expo (ICME), pages 1--6. IEEE, 2020

    2020

  7. [15]

    E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, and W. Chen. Lo RA : Low-rank adaptation of large language models. In International Conference on Learning Representations, 2022

    2022

  8. [16]

    W. Hu, B. Liu, J. Gomes, M. Zitnik, P. Liang, V. S. Pande, and J. Leskovec. Strategies for pre-training graph neural networks. arXiv: Learning, 2019

    2019

  9. [17]

    J. Jeon, J. KIM, H. Song, S. Cho, and N. Park. Gt-gan: General purpose time series synthesis with generative adversarial networks. In NeurIPS, 2022

    2022

  10. [18]

    M. Jin, S. Wang, L. Ma, Z. Chu, J. Y. Zhang, X. L. Shi, P.-Y. Chen, Y. Liang, Y.-F. Li, S. Pan, and Q. Wen. Time-llm: Time series forecasting by reprogramming large language models. ArXiv, abs/2310.01728, 2023

    2023 arXiv

  11. [19]

    X. Jin, Y. Park, D. C. Maddix, B. Wang, and X. Yan. Domain adaptation for time series forecasting via attention sharing. In International Conference on Machine Learning, ICML, 2022

    2022

  12. [20]

    G. Kang, L. Jiang, Y. Yang, and A. Hauptmann. Contrastive adaptation network for unsupervised domain adaptation. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4888--4897, 2019

    2019

  13. [21]

    Karmitsa, S

    N. Karmitsa, S. Taheri, A. M. Bagirov, and P. M \"a kinen. Missing value imputation via clusterwise linear regression. IEEE Transactions on Knowledge and Data Engineering, 34: 0 1889--1901, 2022

    1901

  14. [22]

    Kendall, Y

    A. Kendall, Y. Gal, and R. Cipolla. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In CVPR, 2018

    2018

  15. [23]

    Kitaev, L

    N. Kitaev, L. Kaiser, and A. Levskaya. Reformer: The efficient transformer. ArXiv, abs/2001.04451, 2020 a

    2001 arXiv

  16. [24]

    Kitaev, L

    N. Kitaev, L. Kaiser, and A. Levskaya. Reformer: The efficient transformer. In ICLR, 2020 b

    2020

  17. [25]

    Kudo and J

    T. Kudo and J. Richardson. Sentencepiece: A simple and language independent subword tokenizer and detokenizer for neural text processing. In Conference on Empirical Methods in Natural Language Processing, 2018

    2018

  18. [26]

    Lai, W.-C

    G. Lai, W.-C. Chang, Y. Yang, and H. Liu. Modeling long- and short-term temporal patterns with deep neural networks. The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, 2017

    2017

  19. [27]

    J. Li, G. Li, Y. Shi, and Y. Yu. Cross-domain adaptive clustering for semi-supervised domain adaptation. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 2505--2514, 2021

    2021

  20. [28]

    J. Y. Li, C. Liu, S. Cheng, R. Arcucci, and linda Qiao. Frozen language model helps ecg zero-shot learning. ArXiv, abs/2303.12311, 2023

    2023 arXiv

  21. [29]

    Y. Li, X. Lu, Y. Wang, and D. Dou. Generative time series forecasting with diffusion, denoise, and disentanglement. In NeurIPS, 2022

    2022

  22. [30]

    S. Liu, H. Yu, C. Liao, J. Li, W. Lin, A. X. Liu, and S. Dustdar. Pyraformer: Low-complexity pyramidal attention for long-range time series modeling and forecasting. In ICLR, 2021

    2021

  23. [31]

    X. Liu, J. Hu, Y. Li, S. Diao, Y. Liang, B. Hooi, and R. Zimmermann. Unitime: A language-empowered unified model for cross-domain time series forecasting. ArXiv, abs/2310.09751, 2023 a

    2023 arXiv

  24. [32]

    Y. Liu, H. Wu, J. Wang, and M. Long. Non-stationary transformers: Rethinking the stationarity in time series forecasting. In NeurIPS, 2022

    2022

  25. [33]

    Y. Liu, T. Hu, H. Zhang, H. Wu, S. Wang, L. Ma, and M. Long. itransformer: Inverted transformers are effective for time series forecasting, 2023 b

    2023

  26. [34]

    Mangrulkar, S

    S. Mangrulkar, S. Gugger, L. Debut, Y. Belkada, S. Paul, and B. Bossan. Peft: State-of-the-art parameter-efficient fine-tuning methods. https://github.com/huggingface/peft, 2022

    2022

  27. [35]

    Y. Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers. ArXiv, abs/2211.14730, 2023

    2023 arXiv

  28. [36]

    P. Oza, V. A. Sindagi, V. Vs, and V. M. Patel. Unsupervised domain adaptation of object detectors: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46: 0 4018--4040, 2021

    2021

  29. [37]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, and T. Killeen. Pytorch: An imperative style, high-performance deep learning library. In NeurIPS, 2019

    2019

  30. [38]

    Radford and K

    A. Radford and K. Narasimhan. Improving language understanding by generative pre-training. In ArXiv, 2018

    2018

  31. [39]

    Radford, J

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever. Language models are unsupervised multitask learners. In ArXiv, 2019

    2019

  32. [40]

    Ragab, E

    M. Ragab, E. Eldele, Z. Chen, M. Wu, C. Kwoh, and X. Li. Self-supervised autoregressive domain adaptation for time series data. IEEE Transactions on Neural Networks and Learning Systems, 35: 0 1341--1351, 2021

    2021

  33. [41]

    Ragab, E

    M. Ragab, E. Eldele, W. L. Tan, C.-S. Foo, Z. Chen, M. Wu, C. Kwoh, and X. Li. Adatime: A benchmarking suite for domain adaptation on time series data. ACM Transactions on Knowledge Discovery from Data, 17: 0 1--18, 2022

    2022

  34. [42]

    M. G. Schultz, C. Betancourt, B. Gong, F. Kleinert, M. Langguth, L. H. Leufen, A. Mozaffari, and S. Stadtler. Can deep learning beat numerical weather prediction? Philosophical transactions. Series A, Mathematical, physical, and engineering sciences, 379, 2021

    2021

  35. [43]

    Sennrich, B

    R. Sennrich, B. Haddow, and A. Birch. Neural machine translation of rare words with subword units. ArXiv, abs/1508.07909, 2015

    2015 arXiv

  36. [44]

    Seyfi, J.-F

    A. Seyfi, J.-F. Rajotte, and R. T. Ng. Generating multivariate time series with common source coordinated GAN ( COSCI - GAN ). In NeurIPS, 2022

    2022

  37. [45]

    Shabani, A

    A. Shabani, A. H. S. Abdi, L. Meng, and T. Sylvain. Scaleformer: Iterative multi-scale refining transformers for time series forecasting. ArXiv, 2022

    2022

  38. [46]

    X. Shi, Z. Chen, H. Wang, D. Y. Yeung, W.-K. Wong, and W. chun Woo. Convolutional lstm network: A machine learning approach for precipitation nowcasting. In NIPS, 2015

    2015

  39. [47]

    C. Sun, Y. Li, H. Li, and linda Qiao. Test: Text prototype aligned embedding to activate llm's ability for time series. ArXiv, abs/2308.08241, 2023

    2023 arXiv

  40. [48]

    Touvron, T

    H. Touvron, T. Lavril, G. Izacard, and X. M. et al. Llama: Open and efficient foundation language models. ArXiv, abs/2302.13971, 2023 a

    2023 arXiv

  41. [49]

    Touvron, L

    H. Touvron, L. Martin, K. R. Stone, and e. a. Peter Albert. Llama 2: Open foundation and fine-tuned chat models. ArXiv, abs/2307.09288, 2023 b

    2023 arXiv

  42. [50]

    Tzeng, J

    E. Tzeng, J. Hoffman, K. Saenko, and T. Darrell. Adversarial discriminative domain adaptation. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2962--2971, 2017

    2017

  43. [51]

    van den Oord, O

    A. van den Oord, O. Vinyals, and K. Kavukcuoglu. Neural discrete representation learning. In NIPS, volume abs/1711.00937, 2017

    2017 arXiv

  44. [52]

    Vaswani, N

    A. Vaswani, N. M. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention is all you need. In NeurIPS, 2017

    2017

  45. [53]

    Vibashan, V

    V. Vibashan, V. Gupta, P. Oza, V. A. Sindagi, and V. M. Patel. Mega-cda: Memory guided attention for category-aware unsupervised domain adaptive object detection. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4514--4524, 2021

    2021

  46. [54]

    C. Wang, C. Xu, and D. Tao. Self-supervised pose adaptation for cross-domain image animation. IEEE Transactions on Artificial Intelligence, 1: 0 34--46, 2020 a

    2020

  47. [55]

    H. Wang, H. He, and D. Katabi. Continuously indexed domain adaptation. In ICML, 2020 b

    2020

  48. [56]

    H. Wang, J. Peng, F. Huang, J. Wang, J. Chen, and Y. Xiao. MICN : Multi-scale local and global context modeling for long-term series forecasting. In ICLR, 2023

    2023

  49. [57]

    R. Wang, D. C. Maddix, C. Faloutsos, B. Wang, and R. Yu. Bridging physics-based and data-driven modeling for learning dynamical systems. ArXiv, abs/2011.10616, 2020 c

    2011 arXiv

  50. [58]

    Z. Wang, X. Xu, W. Zhang, G. Trajcevski, T. Zhong, and F. Zhou. Learning latent seasonal-trend representations for time series forecasting. In NeurIPS, 2022 a

    2022

  51. [59]

    Z. Wang, Y. Zhou, R. Wang, T.-Y. Lin, A. Shah, and S. N. Lim. Few-shot fast-adaptive anomaly detection. In Neural Information Processing Systems (NeurIPS), 2022 b

    2022

  52. [60]

    Wilson and D

    G. Wilson and D. J. Cook. A survey of unsupervised deep domain adaptation. ACM Transactions on Intelligent Systems and Technology (TIST), 11: 0 1--46, 2018

    2018

  53. [61]

    G. Woo, C. Liu, D. Sahoo, A. Kumar, and S. Hoi. Co ST : Contrastive learning of disentangled seasonal-trend representations for time series forecasting. In ICLR, 2022 a

    2022

  54. [62]

    G. Woo, C. Liu, D. Sahoo, A. Kumar, and S. C. H. Hoi. Etsformer: Exponential smoothing transformers for time-series forecasting. arXiv preprint arXiv:2202.01381, 2022 b

    2022 arXiv

  55. [63]

    Wu and M

    H. Wu and M. K. Ng. Multiple graphs and low-rank embedding for multi-source heterogeneous domain adaptation. ACM Trans. Knowl. Discov. Data, 16: 0 77:1--77:25, 2022

    2022

  56. [64]

    H. Wu, J. Xu, J. Wang, and M. Long. Autoformer: Decomposition transformers with Auto-Correlation for long-term series forecasting. In NeurIPS, 2021

    2021

  57. [65]

    H. Wu, T. Hu, Y. Liu, H. Zhou, J. Wang, and M. Long. Timesnet: Temporal 2d-variation modeling for general time series analysis. In ICLR, 2023

    2023

  58. [66]

    Z. Wu, S. Pan, G. Long, J. Jiang, X. Chang, and C. Zhang. Connecting the dots: Multivariate time series forecasting with graph neural networks. Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2020

    2020

  59. [67]

    T. Xu, W. Chen, P. Wang, F. Wang, H. Li, and R. Jin. Cdtrans: Cross-domain transformer for unsupervised domain adaptation. In ICLR, 2022

    2022

  60. [68]

    Xue and F

    H. Xue and F. D.Salim. Promptcast: A new prompt-based learning paradigm for time series forecasting. IEEE Transactions on Knowledge and Data Engineering, 2022

    2022

  61. [69]

    K. Yi, Q. Zhang, W. Fan, H. He, L. Hu, P. Wang, N. An, L. Cao, and Z. Niu. Fourier GNN : Rethinking multivariate time series forecasting from a pure graph perspective. In NeurIPS, 2023 a

    2023

  62. [70]

    K. Yi, Q. Zhang, W. Fan, S. Wang, P. Wang, H. He, N. An, D. Lian, L. Cao, and Z. Niu. Frequency-domain MLP s are more effective learners in time series forecasting. In NeurIPS, 2023 b

    2023

  63. [71]

    A. Zeng, M. Chen, L. Zhang, and Q. Xu. Are transformers effective for time series forecasting? In AAAI, 2023

    2023

  64. [72]

    Zhang, D

    Q. Zhang, D. Guo, X. Zhao, L. Yuan, and L. Luo. Discovering frequency bursting patterns in temporal graphs. ICDE, pages 599--611, 2023

    2023

  65. [73]

    Zhang, Y

    T. Zhang, Y. Zhang, W. Cao, J. Bian, X. Yi, S. Zheng, and J. Li. Less is more: Fast multivariate time series forecasting with light sampling-oriented mlp structures. ArXiv, abs/2207.01186, 2022 a

    2022 arXiv

  66. [74]

    Zhang, Z

    X. Zhang, Z. Zhao, T. Tsiligkaridis, and M. Zitnik. Self-supervised contrastive pre-training for time series via time-frequency consistency. In NeurIPS, 2022 b

    2022

  67. [75]

    Zhang, B

    Y. Zhang, B. Deng, H. Tang, L. Zhang, and K. Jia. Unsupervised multi-class domain adaptation: Theory, algorithms, and practice. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44: 0 2775--2792, 2020

    2020

  68. [76]

    S. Zhao, X. Yue, S. Zhang, B. Li, H. Zhao, B. Wu, R. Krishna, J. Gonzalez, A. L. Sangiovanni-Vincentelli, S. A. Seshia, and K. Keutzer. A review of single-source deep unsupervised visual domain adaptation. IEEE Transactions on Neural Networks and Learning Systems, 33: 0 473--493, 2020

    2020

  69. [77]

    H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, and W. Zhang. Informer: Beyond efficient transformer for long sequence time-series forecasting. In AAAI, 2021

    2021

  70. [78]

    T. Zhou, Z. Ma, Q. Wen, X. Wang, L. Sun, and R. Jin. FEDformer : Frequency enhanced decomposed transformer for long-term series forecasting. In ICML, 2022 a

    2022

  71. [79]

    T. Zhou, Z. Ma, xue wang, Q. Wen, L. Sun, T. Yao, W. Yin, and R. Jin. Film: Frequency improved legendre memory model for long-term time series forecasting. In NeurIPS, 2022 b

    2022

  72. [80]

    T. Zhou, P. Niu, X. Wang, L. Sun, and R. Jin. One fits all: Power general time series analysis by pretrained LM . In NIPS, 2023

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.