Pith. sign in

REVIEW 3 major objections 4 minor 39 references

Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning

T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A non-contrastive time series pretraining method, FEI, drops positive and negative pairs and reports higher transfer accuracy on 8 benchmarks.

desk verdict A genuinely new non-contrastive recipe for time series SSL with a clean architecture, but the SOTA claim leans on a fairness premise that the paper doesn't support. read the letter →

arxiv 2412.20790 v2 pith:JRYT3RL4 submitted 2024-12-30 cs.LG cs.AI

En Fu , Yanyan Hu This is my paper · ORCID
classification cs.LGcs.AI
keywords timeseriesrepresentationlearningnon-contrastivefrequencymaskingembeddinginferenceself-supervisedtransfermomentumencoderlinearevaluation
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

Contrastive learning builds its training signal from positive and negative sample pairs, but time series semantics are continuous; a 7-day cycle and a 6.5-day cycle are different without being opposites. The paper's claim is that this mismatch is the root of the field's generalization problems, and that a non-contrastive objective avoids it entirely. Frequency-Masked Embedding Inference (FEI) pretrains an encoder by masking random frequency bands of a series and training it to infer, in embedding space, the representation of the masked series from the mask prompt, and to infer the mask from the representation. On eight classification and regression benchmarks, after one pretraining pass on SleepEEG and direct transfer, FEI reports higher average accuracy and lower error than six contrastive baselines in both linear evaluation and small-sample fine-tuning. If the claim holds, representation learning for time series does not need pair construction at all.

What carries the argument

The carrying object is the paired inference structure around a frequency mask. The original series $x$ is transformed by FFT, a random binary mask $M$ is applied to the amplitude spectrum, and inverse FFT gives a target series $x'$ whose difference from $x$ is continuous in the amount of masking (the paper uses a uniform masking ratio $\beta_1=0.0$ to $\beta_2=0.7$). A mask encoder maps $M$ to $m = M W_{\mathrm{emb}} / \sqrt{k}$, an embedding whose scale stays stable as the number of masked components $k$ varies and which is exactly zero for an empty mask. The original encoder and projector yield embedding $u$; a momentum-updated encoder yields $u'$. One predictor estimates $\hat{u}' = z_{\psi_1}(u + D(m))$ and the other estimates $\hat{m} = z_{\psi_2}(D(u) - u')$, with gradient detachment $D(\cdot)$ separating the two optimization paths. This pair of inverse inference directions forces the encoder to organize embedding space by frequency semantics while preventing the collapse that a single inference direction would invite; the total loss is the sum of two L2 losses.

What would settle it

A decisive test would be to rerun all baselines with their native encoder architectures on the same eight transfer tasks: if the average performance gap between FEI and the best contrastive baseline shrinks to within seed noise or reverses, the claim that FEI generalizes better fails. A quick additional check is to retrain FEI and the strongest baseline with at least five random seeds on a frequency-matched held-out subset; the paper's conclusion predicts FEI should win robustly across seeds.

Watch

Extended reading notes

Core claim

The central discovery is that removing positive and negative pairs does not hurt transferable time series representations; it improves them. FEI creates a target series by masking a random fraction of the amplitude spectrum before inverse Fourier transform, so the semantic distance between original and target varies continuously with how many and which frequency bands are removed. The encoder, a momentum-updated copy, and a mask encoder then support two inverse inference tasks: infer the masked target embedding from the original embedding plus the mask prompt, and infer the mask embedding from the difference between original and target embeddings. The training loss is just the sum of the two L2 distances between inferred and true embeddings, with no negatives and no augmented views. The authors report the best average transfer performance on six classification and two regression datasets whose sampling frequencies range from 1 Hz to 64 kHz, with gains concentrated on frequency-mismatched data.

Load-bearing premise

The load-bearing premise is that forcing every baseline onto the same 1D ResNet encoder (except TimesURL, which could not be swapped due to memory limits) is a fair test of the pretraining objectives; if a baseline's design is coupled to its own architecture, the observed gaps reflect architecture mismatch rather than representation quality.

Editorial extensions

If this is right

  • If FEI's results are correct, time series self-supervision can drop hard-negative mining and augmentation-tuned positive pairs entirely, replacing them with a single frequency-masked inference loss.
  • A pretrained encoder transfers to downstream sampling frequencies from 1 Hz to 64 kHz without per-dataset pretraining, indicating that a frequency-sensitive objective yields broad generalization rather than frequency memorization.
  • The ablations show the mask-inference branch alone prevents representation collapse, which means the framework is not dependent on the momentum encoder for stability.
  • Time-domain masking transfers well only when downstream frequency matches pretraining frequency, while frequency-domain masking stays stable across frequency shifts; this gives users a concrete rule for choosing augmentation type.

Reading between the lines

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

  • The paper leaves implicit that FEI's semantic axis is frequency, so the advantage should be largest on periodic and spectral datasets and smallest on irregular, aperiodic series; this is testable by sorting UCR subsets by spectral coherence.
  • Because only the amplitude spectrum is masked, phase information always survives; a natural extension is phase masking to test whether phase carries independent semantic signal that embedding inference can exploit.
  • The authors' stated future direction suggests continuous modeling at the time-step level; FEI could be adapted to patch or point embeddings for anomaly detection and forecasting, though that extension is not demonstrated in the paper.
  • A sharper validation would measure whether training difficulty tracks the energy in masked frequency bands; if the encoder truly learns frequency semantics, masking high-energy bands should make embedding inference harder and should matter more for downstream transfer.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes Frequency-Masked Embedding Inference (FEI), a non-contrastive self-supervised method for time series representation learning. FEI creates a frequency-masked version of each input series, encodes the original series with a trainable encoder and the masked series with a momentum encoder, and trains two inference branches: one predicts the masked-series embedding from the original embedding plus a mask embedding, and the other predicts the mask embedding from the difference between original and masked embeddings. The pre-training loss is the sum of two L2 terms (Eq. 5). The method is evaluated by pre-training on SleepEEG and transferring to 6 classification datasets plus 128 UCR and 2 regression datasets, under linear evaluation and end-to-end fine-tuning. The paper reports state-of-the-art results against six contrastive baselines and claims that FEI enables continuous semantic modeling without positive/negative sample pairs.

Significance. If the empirical claims are sustained, FEI would be a useful contribution to time series representation learning: it is a simple, non-contrastive objective, it avoids hard-negative construction, and the ablations and visualizations suggest that the frequency-masking prompt mechanism is meaningful. The paper also provides code and reports a fairly broad set of downstream tasks. However, the central state-of-the-art claim is currently not established because the baseline comparison does not hold the encoder architecture constant across methods and because no uncertainty estimates or significance tests are reported. The method itself is coherent and the flaws in the evaluation appear fixable, so the contribution is promising but needs substantial empirical strengthening.

major comments (3)
  1. [Baselines; Tables 2–5] The state-of-the-art claim rests on an uncontrolled encoder comparison. The paper states that 'we use 1-D ResNet as the encoder network for all methods except TimesURL' and explains that TimesURL could not be replaced due to an Out-of-Memory issue, with only its embedding dimension aligned. Several baselines (e.g., TS2Vec, TimeDRL, TF-C, InfoTS) were designed with specific encoder architectures, and their objectives may be coupled to those architectures. Replacing them with a generic 1D ResNet, while evaluating FEI with its own architecture and TimesURL with its native larger encoder, means the observed gaps in Tables 2–5 could reflect architecture mismatch or capacity differences rather than representation quality. No native-encoder comparisons, same-capacity controls, or encoder-sensitivity analyses are provided. This directly undermines the paper's headline claim that FEI 'significantly outperforms existing contrastive-based methods.'
  2. [Task 1: Classification; Task 2: Regression; Tables 2–5] The empirical support for 'significantly outperforms' is weakened by the absence of error bars, confidence intervals, or significance tests. All experiments fix the random seed at '2024' (Appendix) and report single numbers. Several margins are small: for example, linear evaluation on EPI gives 96.84 vs. 96.42 for SimMTM, and on HAR gives 79.54 vs. 79.10 for TimesURL; fine-tuning on EMG gives 97.56, tied with InfoTS. Without repeated runs or a statistical test, these differences are not established as significant, and the average improvements of 2.15% and 3.50% cited in the analysis could be within run-to-run variability.
  3. [Eq. (2); Table 9, Frequency Masking] The definition of the mask embedding is incomplete for the permitted case k = 0. Equation (2) defines m = M W_emb / sqrt(k), but with β1 = 0.0 (Table 9) the masking ratio can be zero, making k = 0 possible; division by sqrt(k) is then undefined. The text states that when M = {0}^n the mask embedding should be 0, but this special case is not incorporated into Eq. (2). The authors should either define the k = 0 convention explicitly or set β1 > 0 to exclude this case, and should confirm that the released implementation handles it consistently.
minor comments (4)
  1. [Appendix, Baselines] The sentence 'we conduct unified pre-training using 1D ResNet as the encoder for various baseline methods, including ... TimesURL, with the exception of TimesURL' is self-contradictory and should be rephrased to match the main-text statement that all methods except TimesURL use the 1D ResNet.
  2. [Table 1; 128 UCR] The paper reports results for '128 UCR' as a single row but does not explain how the 128 constituent datasets are aggregated (e.g., mean accuracy over datasets) or which train/test splits and sample lengths are used; this should be stated explicitly.
  3. [Related Works] The claim that 'JEPA has not yet been applied to time series representation learning' is too strong given that the paper itself cites LaT-PFN (Verdenius et al. 2024), a JEPA-based method for time-series forecasting; the novelty claim should be qualified to representation learning or forecasting excluded.
  4. [Figure 2] The caption of Figure 2 does not identify which lines correspond to 'w/ mask infer.' and 'w/o mask infer.', and the meaning of the 'lowest point' annotations is unclear; please add explicit labels and a legend.

Circularity Check

1 steps flagged · score 4.0 of 10

FD-B headline result is fitted on FD-B via sensitivity sweeps; the FEI inference objective itself is not circular.

  1. fitted input called prediction [Main text Sensitivity paragraph; Appendix Sensitivity (Figure 5); Table 2 FD-B row]
    "Sensitivity We further analyze the impact of different masking ratios on the performance of FEI. Figure 4 shows the linear evaluation accuracy of FEI under different masking ratios on the FD-B dataset. Ultimately, we use β1 = 0.0 and β2 = 0.7. ... The results indicate that the optimal range for α in FEI is between 0.99 and 0.999, within which sufficient representational generalization can be achieved. Ultimately, we use α = 0.995."

    The FD-B row of Table 2 (FEI 67.25%) is produced by a configuration whose two key hyperparameters were selected by inspecting FEI accuracy on the same FD-B benchmark: the masking-ratio bounds β1, β2 from Figure 4 and the momentum factor α from Appendix Figure 5, both reported as accuracy on the FD-B dataset. As written, the paper chooses hyperparameters that perform well on FD-B and then reports that same FD-B performance as evidence of superiority. The FD-B result is therefore the outcome of a selection sweep over FD-B rather than an independent prediction, so it cannot independently support the claim of 'new state-of-the-art performance across 8 benchmark datasets.' The remaining seven datasets were not used for this selection and still provide independent evidence.

full rationale

FEI's training objective (Eq. 5) is a standard JEPA-style self-supervised loss: the mask-encoder and predictors are trained to reconstruct internally generated targets (the momentum-encoder embedding u' and the mask embedding m). These targets are produced by the model itself during pre-training, but that is an internal learning signal, not a derivation of an external result from fitted parameters; the downstream linear-evaluation and fine-tuning numbers are obtained after pre-training with frozen or transferred weights. No load-bearing self-citation or imported uniqueness theorem is used; the JEPA and momentum-encoder citations are external, supporting references rather than circular authority. The only genuinely circular element in the evidence is the FD-B benchmark: both the masking-ratio and momentum hyperparameters are chosen using FD-B accuracy, and FD-B accuracy is then reported as a headline result. That single benchmark result is fitted rather than predicted. Because the central method and the other seven benchmark results remain independent, the overall circularity is partial rather than pervasive.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The method introduces no new physical or mathematical entities. The free parameters are the masking ratio bounds and the momentum factor, both tuned on a downstream benchmark that is also part of the reported results. The axioms are standard domain assumptions about the suitability of L2 embedding distances and the transferability of a single pretraining dataset.

free parameters (2)
  • Masking ratio bounds beta1, beta2 = beta1=0.0, beta2=0.7
    Chosen by sensitivity analysis on FD-B (Figure 4), a dataset that also appears in the main results, so this tuning may inflate FD-B performance.
  • Momentum factor alpha = 0.995
    Selected from sensitivity analysis on FD-B (Figure 5); within the 0.99-0.999 range.
assumptions (3)
  • domain assumption Frequency masking produces semantically continuous variations of the input series.
    The method's premise is that masking different frequency components creates target series whose semantic difference varies continuously with mask position and amount; this is assumed throughout the method design (Section 'Frequency Masking').
  • domain assumption L2 distance in the embedding space is a suitable training signal for semantic relationships.
    The loss in Equation (5) directly compares embeddings with L2; the paper provides no justification that this metric aligns with downstream semantic similarity. This is a standard but unproven assumption.
  • domain assumption The SleepEEG dataset is representative enough to pretrain generalizable time series representations.
    The paper pre-trains on SleepEEG and transfers to various datasets; the success of transfer is an empirical question, but the assumption underpins the experimental design.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning." pith.science (2026). https://pith.science/paper/JRYT3RL4

@misc{pith2026241220790,
  author       = {Pith},
  title        = {Pith review of: Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JRYT3RL4}},
  note         = {Machine review of arXiv:2412.20790}
}
read the original abstract

Contrastive learning underpins most current self-supervised time series representation methods. The strategy for constructing positive and negative sample pairs significantly affects the final representation quality. However, due to the continuous nature of time series semantics, the modeling approach of contrastive learning struggles to accommodate the characteristics of time series data. This results in issues such as difficulties in constructing hard negative samples and the potential introduction of inappropriate biases during positive sample construction. Although some recent works have developed several scientific strategies for constructing positive and negative sample pairs with improved effectiveness, they remain constrained by the contrastive learning framework. To fundamentally overcome the limitations of contrastive learning, this paper introduces Frequency-masked Embedding Inference (FEI), a novel non-contrastive method that completely eliminates the need for positive and negative samples. The proposed FEI constructs 2 inference branches based on a prompting strategy: 1) Using frequency masking as prompts to infer the embedding representation of the target series with missing frequency bands in the embedding space, and 2) Using the target series as prompts to infer its frequency masking embedding. In this way, FEI enables continuous semantic relationship modeling for time series. Experiments on 8 widely used time series datasets for classification and regression tasks, using linear evaluation and end-to-end fine-tuning, show that FEI significantly outperforms existing contrastive-based methods in terms of generalization. This study provides new insights into self-supervised representation learning for time series. The code is available at https://github.com/USTBInnovationPark/Frequency-masked-Embedding-Inference.

Figures

Figures reproduced from arXiv: 2412.20790 by the authors.

Figure 1
Figure 1. The core architecture of FEI consists of 2 sets of [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 1
Figure 1. The overall structure of the proposed FEI. The original time series is fed into the encoder fθ and a sub￾space projector sϕ to generate the original embedding. The target time series is constructed by applying random fre￾quency masking, which is then fed into the momentum en￾coder—a smoothed copy of the original encoder updated via exponential moving average—to produce the target embed￾ding. The goal of FEI is to en… view at source ↗
Figure 2
Figure 2. Comparison of training loss curves for the first 20 [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: Visualization of embedding inference results on [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]
Figure 4
Figure 4. Figure 4: Accuracy results of FEI on the FD-B dataset under [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: The linear evaluation results for varying momen [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Visualization of target series and inference results on the [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Visualization of target series and inference results on the [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 24 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    G.; Lehnertz, K.; Mormann, F.; Rieke, C.; David, P.; and Elger, C

    Andrzejak, R. G.; Lehnertz, K.; Mormann, F.; Rieke, C.; David, P.; and Elger, C. E. 2001. Indications of nonlinear deterministic and finite-dimensional structures in time series of brain electrical activity: Dependence on recording region and brain state. Physical Review E, 64(6): 061907

    work page 2001

  4. [4]

    L.; et al

    Anguita, D.; Ghio, A.; Oneto, L.; Parra, X.; Reyes-Ortiz, J. L.; et al. 2013. A public domain dataset for human activity recognition using smartphones. In Esann, volume 3, 3

    work page 2013

  5. [5]

    Assran, M.; Duval, Q.; Misra, I.; Bojanowski, P.; Vincent, P.; Rabbat, M.; LeCun, Y.; and Ballas, N. 2023. Self-supervised learning from images with a joint-embedding predictive architecture. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15619--15629

    work page 2023

  6. [6]

    Chang, C.; Chan, C.-T.; Wang, W.-Y.; Peng, W.-C.; and Chen, T.-F. 2024. TimeDRL: Disentangled Representation Learning for Multivariate Time-Series. In 2024 IEEE 40th International Conference on Data Engineering (ICDE), 625--638. IEEE

    2024

  7. [7]

    A.; Keogh, E.; Kamgar, K.; Yeh, C.-C

    Dau, H. A.; Keogh, E.; Kamgar, K.; Yeh, C.-C. M.; Zhu, Y.; Gharghabi, S.; Ratanamahatana, C. A.; Yanping; Hu, B.; Begum, N.; Bagnall, A.; Mueen, A.; Batista, G.; and Hexagon-ML. 2018. The UCR Time Series Classification Archive. https://www.cs.ucr.edu/ eamonn/time_series_data_2018/. Accessed: 2024-08-01

    work page 2018

  8. [8]

    Devlin, J.; Chang, M.-W.; Lee, K.; and Toutanova, K. 2019. BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding. In Burstein, J.; Doran, C.; and Solorio, T., eds., Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and ...

    2019

Show all 39 references
  1. [9]

    Dong, J.; Wu, H.; Zhang, H.; Zhang, L.; Wang, J.; and Long, M. 2023. SimMTM: A Simple Pre-Training Framework for Masked Time-Series Modeling. In Oh, A.; Naumann, T.; Globerson, A.; Saenko, K.; Hardt, M.; and Levine, S., eds., Advances in Neural Information Processing Systems, ...

    2023

  2. [10]

    Fang, H.; Wang, S.; Zhou, M.; Ding, J.; and Xie, P. 2020. Cert: Contrastive self-supervised learning for language understanding. arXiv preprint arXiv:2005.12766

    2020 arXiv

  3. [11]

    Gao, Y.; Wen, Y.; and Wu, J. 2021. A Neural Network-Based Joint Prognostic Model for Data Fusion and Remaining Useful Life Prediction. IEEE Transactions on Neural Networks and Learning Systems, 32(1): 117--127

    2021

  4. [12]

    L.; Amaral, L

    Goldberger, A. L.; Amaral, L. A.; Glass, L.; Hausdorff, J. M.; Ivanov, P. C.; Mark, R. G.; Mietus, J. E.; Moody, G. B.; Peng, C.-K.; and Stanley, H. E. 2000. PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. circul...

    2000

  5. [13]

    Grill, J.-B.; Strub, F.; Altch \'e , F.; Tallec, C.; Richemond, P.; Buchatskaya, E.; Doersch, C.; Avila Pires, B.; Guo, Z.; Gheshlaghi Azar, M.; et al. 2020. Bootstrap your own latent-a new approach to self-supervised learning. Advances in neural information processing systems...

    2020

  6. [14]

    Han, K.; Wang, Y.; Chen, H.; Chen, X.; Guo, J.; Liu, Z.; Tang, Y.; Xiao, A.; Xu, C.; Xu, Y.; et al. 2022. A survey on vision transformer. IEEE transactions on pattern analysis and machine intelligence, 45(1): 87--110

    2022

  7. [15]

    H.; Tuk, B.; Kamphuisen, H

    Kemp, B.; Zwinderman, A. H.; Tuk, B.; Kamphuisen, H. A.; and Oberye, J. J. 2000. Analysis of a sleep-dependent neuronal feedback loop: the slow-wave microcontinuity of the EEG. IEEE Transactions on Biomedical Engineering, 47(9): 1185--1194

    2000

  8. [16]

    Lan, X.; Yan, H.; Hong, S.; and Feng, M. 2024. Towards Enhancing Time Series Contrastive Learning: A Dynamic Bad Pair Mining Approach. In The Twelfth International Conference on Learning Representations

    2024

  9. [17]

    LeCun, Y. 2022. A path towards autonomous machine intelligence version 0.9. 2, 2022-06-27. OpenReview

    2022

  10. [18]

    K.; Zimmer, D.; and Sextro, W

    Lessmeier, C.; Kimotho, J. K.; Zimmer, D.; and Sextro, W. 2016. Condition monitoring of bearing damage in electromechanical drive systems by using motor current signals of electric motors: A benchmark data set for data-driven classification. In PHM Society European Conference,...

    2016

  11. [19]

    Li, Z.; Chen, Z.; Yang, F.; Li, W.; Zhu, Y.; Zhao, C.; Deng, R.; Wu, L.; Zhao, R.; Tang, M.; et al. 2021. Mst: Masked self-supervised transformer for visual representation. Advances in Neural Information Processing Systems, 34: 13165--13176

    2021

  12. [20]

    Liu, J.; and Chen, S. 2024. TimesURL: Self-Supervised Contrastive Learning for Universal Time Series Representation Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 38(12): 13918--13926

    2024

  13. [21]

    Liu, J.; Zhong, L.; Wickramasuriya, J.; and Vasudevan, V. 2009. uWave: Accelerometer-based personalized gesture recognition and its applications. Pervasive and Mobile Computing, 5(6): 657--675

    2009

  14. [22]

    Luo, D.; Cheng, W.; Wang, Y.; Xu, D.; Ni, J.; Yu, W.; Zhang, X.; Liu, Y.; Chen, Y.; Chen, H.; and Zhang, X. 2023. Time Series Contrastive Learning with Information-Aware Augmentations. Proceedings of the AAAI Conference on Artificial Intelligence, 37(4): 4534--4542

    2023

  15. [23]

    Mo, S.; and Yun, S. 2024. DMT-JEPA: Discriminative Masked Targets for Joint-Embedding Predictive Architecture. arXiv preprint arXiv:2405.17995

    2024 arXiv

  16. [24]

    Ozyurt, Y.; Feuerriegel, S.; and Zhang, C. 2023. Contrastive Learning for Unsupervised Domain Adaptation of Time Series. In The Eleventh International Conference on Learning Representations

    2023

  17. [25]

    Petzka, H.; Kamp, M.; Adilova, L.; Sminchisescu, C.; and Boley, M. 2021. Relative flatness and generalization. Advances in neural information processing systems, 34: 18420--18432

    2021

  18. [26]

    Saito, A.; and Poovvancheri, J. 2024. Point-JEPA: A Joint Embedding Predictive Architecture for Self-Supervised Learning on Point Cloud. arXiv preprint arXiv:2404.16432

    2024 arXiv

  19. [27]

    Saxena, A.; Goebel, K.; Simon, D.; and Eklund, N. 2008. Damage propagation modeling for aircraft engine run-to-failure simulation. In 2008 international conference on prognostics and health management, 1--9. IEEE

    2008

  20. [28]

    Tonekaboni, S.; Eytan, D.; and Goldenberg, A. 2021. Unsupervised Representation Learning for Time Series with Temporal Neighborhood Coding. In International Conference on Learning Representations

    2021

  21. [29]

    Trirat, P.; Shin, Y.; Kang, J.; Nam, Y.; Na, J.; Bae, M.; Kim, J.; Kim, B.; and Lee, J.-G. 2024. Universal Time-Series Representation Learning: A Survey. arXiv preprint arXiv:2401.03717

    2024 arXiv

  22. [30]

    Van der Maaten, L.; and Hinton, G. 2008. Visualizing data using t-SNE. Journal of machine learning research, 9(11)

    2008

  23. [31]

    Verdenius, S.; Zerio, A.; and Wang, R. L. 2024. LaT-PFN: A Joint Embedding Predictive Architecture for In-context Time-series Forecasting. arXiv preprint arXiv:2405.10093

    2024 arXiv

  24. [32]

    Wang, B.; Lei, Y.; Li, N.; and Li, N. 2018. A hybrid prognostics approach for estimating remaining useful life of rolling element bearings. IEEE Transactions on Reliability, 69(1): 401--412

    2018

  25. [33]

    Wang, L.; Zhu, Z.; and Zhao, X. 2024. Dynamic predictive maintenance strategy for system remaining useful life prediction via deep learning ensemble method. Reliability Engineering & System Safety, 245: 110012

    2024

  26. [34]

    Wang, Y.; Han, Y.; Wang, H.; and Zhang, X. 2023. Contrast Everything: A Hierarchical Contrastive Framework for Medical Time-Series. In Oh, A.; Naumann, T.; Globerson, A.; Saenko, K.; Hardt, M.; and Levine, S., eds., Advances in Neural Information Processing Systems, volume 36,...

    2023

  27. [35]

    Xu, M.; Moreno, A.; Wei, H.; Marlin, B.; and Rehg, J. M. 2024. REBAR : Retrieval-Based Reconstruction for Time-series Contrastive Learning. In The Twelfth International Conference on Learning Representations

    2024

  28. [36]

    Yan, Y.; Li, R.; Wang, S.; Zhang, F.; Wu, W.; and Xu, W. 2021. ConSERT: A Contrastive Framework for Self-Supervised Sentence Representation Transfer. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Con...

    2021

  29. [37]

    Yue, Z.; Wang, Y.; Duan, J.; Yang, T.; Huang, C.; Tong, Y.; and Xu, B. 2022. TS2Vec: Towards Universal Representation of Time Series. Proceedings of the AAAI Conference on Artificial Intelligence, 36(8): 8980--8987

    2022

  30. [38]

    Zerveas, G.; Jayaraman, S.; Patel, D.; Bhamidipaty, A.; and Eickhoff, C. 2021. A transformer-based framework for multivariate time series representation learning. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 2114--2124

    2021

  31. [39]

    Zhang, X.; Zhao, Z.; Tsiligkaridis, T.; and Zitnik, M. 2022. Self-supervised contrastive pre-training for time series via time-frequency consistency. Advances in Neural Information Processing Systems, 35: 3988--4003

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.