Pith. sign in

REVIEW 5 major objections 5 minor 41 references

FreSH: Frequency-Segmented Hierarchical Multi-Expert Framework for Multivariate Time Series Classification

T0 review · 5 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 FreSH splits the Fourier spectrum into expert-handled bands and reports 76.1% average accuracy on 30 UEA multivariate time series benchmarks with only 54,243 parameters on a vibration test.

desk verdict FreSH's frequency-segmented expert idea is worth a look, but the missing complex-to-real specification and test-set hyperparameter tuning undercut the headline numbers. read the letter →

arxiv 2608.08207 v1 pith:HZZVYHFK submitted 2026-08-08 cs.LG cs.AI

Pingping Liu , Muyao Wang , Zijian Zhang , Tongshun Zhang , Hao Miao , Guorui Xie , Qingliang Li , Qiuzhan Zhou This is my paper · ORCID
classification cs.LGcs.AI
keywords multivariatetimeseriesclassificationfrequency-domainanalysisspectralsegmentationmixture-of-expertsadaptivegatingpolynomiallossclassimbalanceUEAbenchmark
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

FreSH is a multivariate time series classifier that works almost entirely in the frequency domain. It takes the FFT of each input, splits the spectrum into a small number of bands, and assigns each band a set of lightweight MLP experts, with additional full-spectrum experts supplying global context. An adaptive gating layer learns which bands and experts matter for each sample, and a simplified polynomial loss replaces cross-entropy to cope with class imbalance and sample difficulty. The paper reports 76.1% average accuracy and a 3.2 average rank across 30 UEA benchmark datasets, and 94.37% accuracy with only 54,243 parameters on a real-world vibration dataset. The claim is that band-specific specialization plus cheap experts captures multi-scale temporal patterns that holistic time- or frequency-domain models miss.

What carries the argument

The carrying object is the frequency-segmented hierarchical mixture-of-experts module. After an FFT maps a sample to complex coefficients $F\in\mathbb{C}^{d\times s}$, the spectrum is padded and split into $K$ equal-length segments $F^{(k)}\in\mathbb{C}^{d\times l_k}$; each segment is processed by $M$ local MLP experts whose outputs are averaged, while $N$ global MLP experts process the full spectrum under a softmax gate $\lambda$, a segment gate $\beta$ reweights the local segment outputs, and a final gate $\hat{\alpha}$ with scalar $\gamma$ fuses the two paths via $H^*=F+H_{\text{local}}+\gamma\,\hat{\alpha}\,H_{\text{global}}$. The inverse FFT of $H^*$ is then passed to a linear classifier. This design keeps every operation except the FFT linear in sequence length, giving complexity $O(d n \log n + m h n + g h n)$, which is the basis of the efficiency claims. The second carrying object is the P-Loss objective $L_P=-\frac{1}{N}\sum_i \log \hat{y}_i + \lambda_P \frac{1}{N}\sum_i (1-\hat{y}_i)^2$, the simplified polynomial loss that replaces cross-entropy.

What would settle it

Re-run FreSH on a fixed UEA dataset, for example AtrialFibrillation where the paper reports 73.3% accuracy, changing only the complex-to-real conversion of $F^{(k)}$: magnitude-only versus phase-only versus stacked real/imaginary inputs. If the accuracies are nearly identical, phase information plays no role and the frequency-segmentation story is weaker than claimed; if they differ by several points, the published numbers depend on a preprocessing detail the paper does not document.

Watch

Extended reading notes

Core claim

The paper's central claim is that the bottleneck in multivariate time series classification is frequency selectivity, not model capacity. It argues that models treating the spectrum as one holistic object cannot separate band-specific patterns, and that time-domain models miss periodic structure; FreSH therefore divides the complex FFT spectrum into $K$ equal-length bands, processes each band with its own averaged ensemble of small MLP experts, and adds gated full-spectrum experts whose output is fused through $H^* = F + H_{\text{local}} + \gamma \hat{\alpha} H_{\text{global}}$ before an inverse FFT and a linear classifier. The same segmentation thesis extends to the loss: a second-order polynomial correction $(1 - \hat{y}_i)^2$ added to the negative log-likelihood is, in the paper's ablations, the single largest contributor to accuracy, and removing it drops average accuracy from 76.1% to 72.9%. On 30 UEA datasets FreSH reports the highest average accuracy (76.1%) and lowest average rank (3.2) among the 10 compared MTSC baselines, and on a vibration dataset it reports 94.37% accuracy at 1.2 ms per batch.

Load-bearing premise

The load-bearing premise is that complex-valued Fourier segments can be fed into the paper's real-valued MLP experts, yet the paper never says how the complex numbers are converted to real inputs; if that conversion discards phase information or misaligns dimensions, the entire learned representation rests on a choice the paper leaves unspecified.

Editorial extensions

If this is right

  • If the reported numbers hold, a frequency-segmented expert model can beat dedicated MTSC models and Transformer-based series models on the UEA benchmark without using global attention at all.
  • The ablation results imply that the choice of loss is as important as the architecture: replacing P-Loss with cross-entropy costs 3.2 points of average accuracy, the largest single drop in the study.
  • The efficiency comparison implies that a 54,243-parameter model can reach 94.37% accuracy on vibration data where baselines with up to hundreds of millions of parameters are slower and less accurate, which supports latency-constrained deployment.
  • The complexity bound $O(d n \log n + m h n + g h n)$ implies the framework scales to long multivariate series, since expert and gating costs grow only linearly with sequence length once the FFT is computed.

Reading between the lines

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

  • The paper never states how the complex-valued segments $F^{(k)}$ are converted to inputs for the real-valued MLP experts; magnitude-only, phase-only, and stacked real/imaginary inputs are all consistent with the equations as written, and each would change what the experts can learn, so the published gains are tied to an undocumented preprocessing choice.
  • Equal-length frequency bands are a convenience rather than a learned design; replacing them with energy-adaptive or learnable band boundaries is a direct extension that could matter on datasets with uneven spectral energy, such as AtrialFibrillation and PhonemeSpectra, where FreSH's accuracy varies sharply.
  • If P-Loss is the largest ablation contributor, the same second-order correction should transfer to other MTSC backbones; this is not claimed in the paper, but the paper's own ablation makes it the most obvious next experiment.
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

5 major / 5 minor

Summary. The paper proposes FreSH, a frequency-domain hierarchical mixture-of-experts framework for multivariate time series classification. The input is transformed with an FFT, the spectrum is split into K segments, local MLP experts process each segment, global MLP experts process the full spectrum, and an adaptive gating mechanism combines the local and global representations. The final representation is mapped back to the time domain with an inverse FFT and classified by a linear layer. Training uses a modified polynomial loss ('P-Loss') and an adaptive mixup schedule. The authors report experiments on 30 UEA datasets, a 10-dataset comparison, ablations, a hyperparameter study, and an efficiency comparison on a vibration dataset, claiming an average accuracy of 76.1%, an average rank of 3.2, and 54,243 parameters.

Significance. If the reported results are reproducible and statistically sound, the frequency-segmented multi-expert design is a plausible lightweight alternative to transformer-based MTSC models, and the efficiency numbers in Table 4 are attractive. The paper has several concrete strengths: the code is publicly available, the evaluation spans 30 UEA datasets plus a real-world dataset, ablations are provided for each major component, and the complexity analysis in Section 2.8 is a useful addition. However, the current text does not fully specify the forward pass, because complex Fourier coefficients are consumed by real-valued MLPs and gates without any stated complex-to-real conversion, and the inverse FFT step is not guaranteed to produce a real signal. In addition, the reported averages lack error bars and significance tests, and key hyperparameters appear to be selected on the test sets. These issues must be resolved before the central quantitative claims can be accepted.

major comments (5)
  1. [§2.5, Eqs. (3)–(10)] The forward pass is underspecified in a load-bearing way. After Eq. (1), F and each segment F^(k) are complex-valued (C^{d×s_padded}), yet the local experts E_m^(k), the global experts U_i, the gate σ_global, and the gate σ_segment are all described as MLPs or linear layers followed by softmax, which conventionally operate on real inputs. The paper never states whether the model uses magnitude, phase, real/imaginary stacking, or complex-valued linear layers, nor how the d×l_k segment is flattened or otherwise fed into each MLP. This choice changes the input dimensionality and the information available to every expert and gate, so the reported 76.1% accuracy, the ablation numbers, and the parameter count are not reproducible from the text alone. Please specify the exact complex-to-real conversion, the input shape of every MLP, and, if complex linear layers are used, the corresponding parameter-count accounting.
  2. [§2.6, Eq. (11)] Equation (11) applies an inverse FFT to H*, which is complex by construction, and then Eq. (12) feeds the resulting X* to a real-valued fully connected layer. An unconstrained inverse FFT of a complex spectrum does not generally produce a real-valued time-domain signal, so the text must specify that a real-only inverse transform (e.g., torch.fft.irfft) or an explicit conjugate-symmetry constraint is applied. Without this, the dimensions and even the real-valuedness of the classifier input are undefined.
  3. [§3.2, Table 1] There is an internal inconsistency in the reported average rank. The text says FreSH 'achieves the highest average accuracy of 76.1% and the best (lowest) average rank of 3.4 among all compared methods,' but the next sentence says it surpasses FreRA by 0.6 in average ranking '(3.2 vs. 3.8),' and Table 1 reports an average rank of 3.2. Please correct the text and table so that all occurrences agree.
  4. [§3.5, Figure 3; §3.1] The hyperparameter analysis appears to select K (number of segments), M (experts per segment), and the number of global experts by comparing average accuracy on the same 30 UEA test datasets used in Table 1, with no validation split. If that is the case, the reported test-set numbers are optimistically biased and cannot be interpreted as an unbiased comparison with baselines whose hyperparameters were not tuned in the same way. Please state explicitly whether any held-out validation data were used, report the selection protocol, and, ideally, provide standard deviations over multiple seeds and a significance test for the main comparisons.
  5. [§3.1] The experimental setup section does not specify the optimizer, learning rate, batch size, number of epochs, or the adaptive mixup reduction schedule, and Section 2.7 does not give the value of λ_P. These are essential for reproducibility, especially because the loss and augmentation schedule are part of the proposed method. Please add a full training configuration table or refer precisely to the configuration in the released code.
minor comments (5)
  1. [Table 4] The DLinear row appears malformed: '#Params' and 'Batch (ms)' are merged as '3,927,0030.66'. Please separate the parameter count from the latency value.
  2. [Table 2] The dataset names 'SelfReglationSCP1' and 'SelfReglationSCP2' are misspelled; they should be 'SelfRegulationSCP1' and 'SelfRegulationSCP2'.
  3. [Abstract] The phrase 'substantially reducing model size and efficiency' is logically reversed; reducing efficiency would be a drawback. It should be 'reducing model size and improving efficiency' or 'reducing model size and computational cost'.
  4. [§2.3] The notation s_padded is introduced without a precise definition of how it relates to the original sequence length l and to the FFT size. Please define s_padded and state whether zero-padding is applied before or after the FFT, since this affects the frequency resolution and the dimensions in Eqs. (2)–(11).
  5. [§2.7, Eq. (13)] The notation for the predicted probability is inconsistent: the text uses 'where ˆy i represents the predicted probability of the true class' but the equation writes 'log ˆyi' without bold or explicit indexing. Please use a consistent notation such as \(\hat{y}_i\) and state that the sum runs over the batch or dataset.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: FreSH's derivation chain is self-contained, and the reported issues are reproducibility or benchmark-tuning concerns rather than circular reductions.

full rationale

FreSH's methodology is a standard supervised learning pipeline: the input is transformed by FFT (Eq. 1), split into frequency segments (Eq. 2), passed through segment-wise and global experts (Eqs. 3-7), adaptively gated (Eqs. 8-9), combined (Eq. 10), transformed back by inverse FFT (Eq. 11), and classified (Eq. 12). Every learned quantity is a function of the original input F, and no predicted output is defined in terms of the reported accuracy or rank. The P-Loss in Eq. (13) is explicitly adapted from Leng et al. (2022) with attribution, and mixup is taken from Zhang et al. (2017); neither is presented as a first-principles derivation that could collapse into its own inputs. The UEA benchmarks are external, and no load-bearing self-citation or imported uniqueness theorem appears. The two substantive concerns are non-circular: (i) the complex-valued FFT output is fed to MLPs and gates without a stated complex-to-real conversion in Eqs. (3), (5), (6), and (8), which makes the method under-specified and hard to reproduce, and (ii) Section 3.5 selects hyperparameters (number of segments and experts) using average accuracy on the same 30 UEA datasets, which may inflate reported performance. These are reproducibility and evaluation-bias issues, not cases where the derivation reduces by construction to its inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claims rest on standard Fourier-transform algebra and on several domain assumptions about frequency segmentation, expert ensembling, and loss design that the paper adopts without independent verification. The key free parameters are the architecture hyperparameters and the loss weight, none of which are given a principled selection procedure; the most consequential ones were tuned on the test sets.

free parameters (6)
  • K (number of frequency segments) = 3
    Chosen via hyperparameter analysis on the 30 UEA test sets (Figure 3a); affects the granularity of frequency decomposition and is load-bearing for the architecture.
  • M (experts per segment) = 3
    Chosen via hyperparameter analysis on the 30 UEA test sets (Figure 3b); affects ensemble strength and overfitting risk.
  • N (number of global experts) = not specified in text
    The paper mentions N global experts but never gives the value used in experiments; the global expert count is a free design choice that affects capacity and gating.
  • lambda_P (P-Loss weight) = not specified
    The second-order term weight in Eq. (13) is a hyperparameter that controls the balance between cross-entropy and the polynomial penalty; no value or tuning procedure is reported.
  • adaptive mixup reduction schedule = not specified
    The strategy reduces the mixing ratio when loss stagnates, but the threshold and reduction amount are not given, making the augmentation schedule an unstated hand-chosen element.
  • training hyperparameters (learning rate, epochs, batch size, optimizer) = not specified
    No training configuration is reported anywhere, so the optimization schedule is effectively a set of unstated hand-chosen parameters on which the results depend.
assumptions (5)
  • standard math FFT and inverse FFT preserve all information in the time series and are invertible without loss.
    Invoked in Section 2.3 Eq. (1) and Section 2.6 Eq. (11); standard property of Fourier transforms.
  • domain assumption Mixup augmentation improves generalization of the classifier.
    Adopted from Zhang et al. 2017 without re-validation; Section 2.3.
  • domain assumption Frequency-domain representation and equal-length segmentation retain class-discriminative information.
    The paper states that segmentation 'allows specialized experts to focus on specific bands' (Section 2.4), but provides no analysis of information loss or why equal-length banding is appropriate across diverse datasets.
  • domain assumption Averaging local expert outputs is at least as good as gating within a segment.
    Section 2.5 states 'avoiding intra-segment gating while still allowing for ensemble effects among experts' without empirical comparison to intra-segment gating.
  • domain assumption The second-order polynomial loss term improves class-imbalance handling compared to cross-entropy and focal loss.
    Borrowed from PolyLoss (Leng et al. 2022) and asserted for MTSC in Section 2.7; no theoretical or empirical isolation beyond one ablation on average accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FreSH: Frequency-Segmented Hierarchical Multi-Expert Framework for Multivariate Time Series Classification." pith.science (2026). https://pith.science/paper/HZZVYHFK

@misc{pith2026260808207,
  author       = {Pith},
  title        = {Pith review of: FreSH: Frequency-Segmented Hierarchical Multi-Expert Framework for Multivariate Time Series Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HZZVYHFK}},
  note         = {Machine review of arXiv:2608.08207}
}
read the original abstract

Multivariate Time Series Classification (MTSC) demands models that can effectively capture complex temporal patterns across multiple scales while remaining computationally efficient. However, existing approaches generally struggle to reconcile fine-grained representation learning, especially under class imbalance and real-world constraints. In this paper, we present FreSH, a Frequency-Segmented Hierarchical Multi-Expert Framework designed to address these challenges. FreSH introduces a new perspective for MTSC by enabling adaptive, multi-scale analysis of temporal signals, allowing different aspects of the data to be modeled in a complementary and coordinated manner. By combining localized specialization with holistic context modeling, FreSH achieves strong representational capacity without incurring excessive computational overhead. An adaptive fusion strategy further enhances flexibility, enabling the model to dynamically emphasize the most informative components of the input. In addition, we incorporate a more robust optimization objective that improves learning stability across varying sample difficulties and class distributions. Extensive evaluations on 30 UEA benchmark datasets and real-world vibration data demonstrate that FreSH consistently outperforms state-of-the-art methods in classification accuracy, while substantially reducing model size and efficiency.

Figures

Figures reproduced from arXiv: 2608.08207 by the authors.

Figure 1
Figure 1. Some models in MTSC cannot adapt to the diverse tem [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Framework overview of FreSH. After transforming time [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Experiments on 30 UEA datasets to evaluate different [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 24 canonical work pages

  1. [1]

    A comprehensive review on machine learning in healthcare industry: classification, restrictions, opportunities and challenges.Sensors, 23(9):4178,

    [Anet al., 2023 ] Qi An, Saifur Rahman, Jingwen Zhou, and James Jin Kang. A comprehensive review on machine learning in healthcare industry: classification, restrictions, opportunities and challenges.Sensors, 23(9):4178,

    work page 2023

  2. [6]

    Efficiently modeling long sequences with structured state spaces.arXiv preprint arXiv:2111.00396,

    [Guet al., 2021 ] Albert Gu, Karan Goel, and Christopher R´e. Efficiently modeling long sequences with structured state spaces.arXiv preprint arXiv:2111.00396,

    arXiv 2021

  3. [11]

    Modeling long-and short-term temporal patterns with deep neural networks

    [Laiet al., 2018 ] Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. Modeling long-and short-term temporal patterns with deep neural networks. InThe 41st international ACM SIGIR conference on research & devel- opment in information retrieval, pages 95–104,

    2018

  4. [12]

    Polyloss: A polynomial expansion perspective of classification loss functions.arXiv preprint arXiv:2204.12511,

    [Lenget al., 2022 ] Zhaoqi Leng, Mingxing Tan, Chenxi Liu, Ekin Dogus Cubuk, Xiaojie Shi, Shuyang Cheng, and Dragomir Anguelov. Polyloss: A polynomial expansion perspective of classification loss functions.arXiv preprint arXiv:2204.12511,

    arXiv 2022

  5. [14]

    Human activity recognition based on multienvironment sensor data.Information Fusion, 91:47– 63,

    [Liet al., 2023 ] Yang Li, Guanci Yang, Zhidong Su, Shaobo Li, and Yang Wang. Human activity recognition based on multienvironment sensor data.Information Fusion, 91:47– 63,

    work page 2023

  6. [15]

    Focal loss for dense ob- ject detection

    [Linet al., 2017 ] Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Doll ´ar. Focal loss for dense ob- ject detection. InProceedings of the IEEE international conference on computer vision, pages 2980–2988,

    work page 2017

  7. [16]

    Scinet: Time series modeling and forecasting with sample convo- lution and interaction.Advances in Neural Information Processing Systems, 35:5816–5828,

    [Liuet al., 2022 ] Minhao Liu, Ailing Zeng, Muxi Chen, Zhi- jian Xu, Qiuxia Lai, Lingna Ma, and Qiang Xu. Scinet: Time series modeling and forecasting with sample convo- lution and interaction.Advances in Neural Information Processing Systems, 35:5816–5828,

    2022

  8. [17]

    Mod- erntcn: A modern pure convolution structure for general time series analysis

    [Luo and Wang, 2024] Donghao Luo and Xue Wang. Mod- erntcn: A modern pure convolution structure for general time series analysis. InThe twelfth international confer- ence on learning representations, pages 1–43,

    work page 2024

Show all 41 references
  1. [18]

    Time series contrastive learning with information-aware augmentations

    [Luoet al., 2023 ] Dongsheng Luo, Wei Cheng, Yingheng Wang, Dongkuan Xu, Jingchao Ni, Wenchao Yu, Xuchao Zhang, Yanchi Liu, Yuncong Chen, Haifeng Chen, et al. Time series contrastive learning with information-aware augmentations. InProceedings of the AAAI Conference on Artific...

    2023

  2. [19]

    Mptsnet: Integrating multiscale peri- odic local patterns and global dependencies for multivari- ate time series classification

    [Muet al., 2025 ] Yang Mu, Muhammad Shahzad, and Xiao Xiang Zhu. Mptsnet: Integrating multiscale peri- odic local patterns and global dependencies for multivari- ate time series classification. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 1...

    2025

  3. [21]

    [Ruizet al., 2021 ] Alejandro Pasos Ruiz, Michael Flynn, James Large, Matthew Middlehurst, and Anthony Bag- nall. The great multivariate time series classification bake off: a review and experimental evaluation of recent algo- rithmic advances.Data mining and knowledge discove...

    2021

  4. [22]

    Omni- scale cnns: a simple and effective kernel size config- uration for time series classification.arXiv preprint arXiv:2002.10061,

    [Tanget al., 2020 ] Wensi Tang, Guodong Long, Lu Liu, Tianyi Zhou, Michael Blumenstein, and Jing Jiang. Omni- scale cnns: a simple and effective kernel size config- uration for time series classification.arXiv preprint arXiv:2002.10061,

    2020 arXiv

  5. [24]

    Frera: A frequency-refined augmentation for contrastive learning on time series classification.arXiv preprint arXiv:2505.23181,

    [Tianet al., 2025 ] Tian Tian, Chunyan Miao, and Hang- wei Qian. Frera: A frequency-refined augmentation for contrastive learning on time series classification.arXiv preprint arXiv:2505.23181,

    2025 arXiv

  6. [25]

    Unsupervised representation learn- ing for time series with temporal neighborhood coding

    [Tonekaboniet al., 2021 ] Sana Tonekaboni, Danny Eytan, and Anna Goldenberg. Unsupervised representation learn- ing for time series with temporal neighborhood coding. arXiv preprint arXiv:2106.00750,

    2021 arXiv

  7. [26]

    Time series classification from scratch with deep neural networks: A strong baseline

    [Wanget al., 2017 ] Zhiguang Wang, Weizhong Yan, and Tim Oates. Time series classification from scratch with deep neural networks: A strong baseline. In2017 Interna- tional joint conference on neural networks (IJCNN), pages 1578–1585. IEEE,

    2017

  8. [27]

    Micn: Multi-scale local and global context modeling for long- term series forecasting

    [Wanget al., 2023 ] Huiqiang Wang, Jian Peng, Feihu Huang, Jince Wang, Junhui Chen, and Yifei Xiao. Micn: Multi-scale local and global context modeling for long- term series forecasting. InThe eleventh international con- ference on learning representations,

    2023

  9. [28]

    Transformers in time series: A survey.arXiv preprint arXiv:2202.07125,

    [Wenet al., 2022 ] Qingsong Wen, Tian Zhou, Chaoli Zhang, Weiqi Chen, Ziqing Ma, Junchi Yan, and Liang Sun. Transformers in time series: A survey.arXiv preprint arXiv:2202.07125,

    2022 arXiv

  10. [29]

    Autoformer: Decomposition transform- ers with auto-correlation for long-term series forecast- ing.Advances in neural information processing systems, 34:22419–22430,

    [Wuet al., 2021 ] Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. Autoformer: Decomposition transform- ers with auto-correlation for long-term series forecast- ing.Advances in neural information processing systems, 34:22419–22430,

    2021

  11. [30]

    Timesnet: Temporal 2d-variation modeling for general time series analysis.arXiv preprint arXiv:2210.02186,

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

    arXiv

  12. [31]

    Flowformer: Lineariz- ing transformers with conservation flows.arXiv preprint arXiv:2202.06258,

    [Wuet al., 2022b ] Haixu Wu, Jialong Wu, Jiehui Xu, Jian- min Wang, and Mingsheng Long. Flowformer: Lineariz- ing transformers with conservation flows.arXiv preprint arXiv:2202.06258,

    arXiv

  13. [32]

    Frequency-domain mlps are more effective learners in time series forecasting

    [Yiet al., 2023 ] Kun Yi, Qi Zhang, Wei Fan, Shoujin Wang, Pengyang Wang, Hui He, Ning An, Defu Lian, Long- bing Cao, and Zhendong Niu. Frequency-domain mlps are more effective learners in time series forecasting. Advances in Neural Information Processing Systems, 36:76656–76679,

    2023

  14. [33]

    Ts2vec: Towards universal representation of time series

    [Yueet al., 2022 ] Zhihan Yue, Yujing Wang, Juanyong Duan, Tianmeng Yang, Congrui Huang, Yunhai Tong, and Bixiong Xu. Ts2vec: Towards universal representation of time series. InProceedings of the AAAI conference on ar- tificial intelligence, volume 36, pages 8980–8987,

    2022

  15. [34]

    Are transformers effective for time series fore- casting? InProceedings of the AAAI conference on artifi- cial intelligence, volume 37, pages 11121–11128,

    [Zenget al., 2023 ] Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu. Are transformers effective for time series fore- casting? InProceedings of the AAAI conference on artifi- cial intelligence, volume 37, pages 11121–11128,

    2023

  16. [35]

    Crossformer: Transformer utilizing cross-dimension de- pendency for multivariate time series forecasting

    [Zhang and Yan, 2023] Yunhao Zhang and Junchi Yan. Crossformer: Transformer utilizing cross-dimension de- pendency for multivariate time series forecasting. InThe eleventh international conference on learning representa- tions,

    2023

  17. [36]

    mixup: Be- yond empirical risk minimization.arXiv preprint arXiv:1710.09412,

    [Zhanget al., 2017 ] Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Be- yond empirical risk minimization.arXiv preprint arXiv:1710.09412,

    2017 arXiv

  18. [37]

    Tapnet: Multivariate time se- ries classification with attentional prototypical network

    [Zhanget al., 2020 ] Xuchao Zhang, Yifeng Gao, Jessica Lin, and Chang-Tien Lu. Tapnet: Multivariate time se- ries classification with attentional prototypical network. In Proceedings of the AAAI conference on artificial intelli- gence, volume 34, pages 6845–6852,

    2020

  19. [38]

    Time series classification using multi-channels deep convolutional neural networks

    [Zhenget al., 2014 ] Yi Zheng, Qi Liu, Enhong Chen, Yong Ge, and J Leon Zhao. Time series classification using multi-channels deep convolutional neural networks. InIn- ternational conference on web-age information manage- ment, pages 298–310. Springer,

    2014

  20. [40]

    Informer: Beyond efficient transformer for long sequence time-series forecasting

    [Zhouet al., 2021 ] Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wan- cai Zhang. Informer: Beyond efficient transformer for long sequence time-series forecasting. InProceedings of the AAAI conference on artificial intelligence, volume 35, pa...

    2021

  21. [41]

    Svp-t: A shape-level variable-position transformer for multivariate time series classification

    [Zuoet al., 2023 ] Rundong Zuo, Guozhong Li, Byron Choi, Sourav S Bhowmick, Daphne Ngar-yin Mah, and Grace LH Wong. Svp-t: A shape-level variable-position transformer for multivariate time series classification. In Proceedings of the AAAI Conference on Artificial Intelli- genc...

    2023

  22. [2014]

    Auto tcl: Automated time series contrastive learning with adaptive augmentations

    [Zhenget al., 2023 ] Xu Zheng, Tianchun Wang, Wei Cheng, Aitian Ma, Haifeng Chen, Mo Sha, and Dongsheng Luo. Auto tcl: Automated time series contrastive learning with adaptive augmentations. InProc. 32nd Int. Joint Conf. Ar- tif. Intell.(IJCAI), pages 1–19,

    2023

  23. [2015]

    Deep learning for time series classi- fication: a review.Data mining and knowledge discovery, 33(4):917–963,

    [Ismail Fawazet al., 2019 ] Hassan Ismail Fawaz, Germain Forestier, Jonathan Weber, Lhassane Idoumghar, and Pierre-Alain Muller. Deep learning for time series classi- fication: a review.Data mining and knowledge discovery, 33(4):917–963,

    2019

  24. [2016]

    Time-series pattern recognition in smart manufacturing systems: A literature review and ontology.Journal of Manufacturing Systems, 69:208–241,

    [Farahaniet al., 2023 ] Mojtaba A Farahani, MR Mc- Cormick, Robert Gianinny, Frank Hudacheck, Ramy Harik, Zhichao Liu, and Thorsten Wuest. Time-series pattern recognition in smart manufacturing systems: A literature review and ontology.Journal of Manufacturing Systems, 69:208–241,

    2023

  25. [2017]

    Multivariate lstm-fcns for time series classification.Neural networks, 116:237–245,

    [Karimet al., 2019 ] Fazle Karim, Somshubra Majumdar, Houshang Darabi, and Samuel Harford. Multivariate lstm-fcns for time series classification.Neural networks, 116:237–245,

    2019

  26. [2018]

    Msgnet: Learning multi- scale inter-series correlations for multivariate time series forecasting

    [Caiet al., 2024 ] Wanlin Cai, Yuxuan Liang, Xianggen Liu, Jianshuai Feng, and Yuankai Wu. Msgnet: Learning multi- scale inter-series correlations for multivariate time series forecasting. InProceedings of the AAAI conference on ar- tificial intelligence, volume 38, pages 11141–11149,

    2024

  27. [2019]

    Lstm fully convolu- tional networks for time series classification.IEEE access, 6:1662–1669,

    [Karimet al., 2017 ] Fazle Karim, Somshubra Majumdar, Houshang Darabi, and Shun Chen. Lstm fully convolu- tional networks for time series classification.IEEE access, 6:1662–1669,

    2017

  28. [2020]

    What makes for good views for contrastive learning?Advances in neural information processing systems, 33:6827–6839,

    [Tianet al., 2020 ] Yonglong Tian, Chen Sun, Ben Poole, Dilip Krishnan, Cordelia Schmid, and Phillip Isola. What makes for good views for contrastive learning?Advances in neural information processing systems, 33:6827–6839,

    2020

  29. [2021]

    Early clas- sification on multivariate time series.Neurocomputing, 149:777–787,

    [Heet al., 2015 ] Guoliang He, Yong Duan, Rong Peng, Xi- aoyuan Jing, Tieyun Qian, and Lingling Wang. Early clas- sification on multivariate time series.Neurocomputing, 149:777–787,

    2015

  30. [2022]

    Shapenet: A shapelet-neural network ap- proach for multivariate time series classification

    [Liet al., 2021 ] Guozhong Li, Byron Choi, Jianliang Xu, Sourav S Bhowmick, Kwok-Pan Chun, and Grace Lai- Hung Wong. Shapenet: A shapelet-neural network ap- proach for multivariate time series classification. InPro- ceedings of the AAAI conference on artificial intelligence, v...

    2021

  31. [2023]

    The uea multivari- ate time series classification archive, 2018.arXiv preprint arXiv:1811.00075,

    [Bagnallet al., 2018 ] Anthony Bagnall, Hoang Anh Dau, Ja- son Lines, Michael Flynn, James Large, Aaron Bostrom, Paul Southam, and Eamonn Keogh. The uea multivari- ate time series classification archive, 2018.arXiv preprint arXiv:1811.00075,

    2018 arXiv

  32. [2024]

    Multi-scale convolutional neural networks for time series classification.arXiv preprint arXiv:1603.06995,

    [Cuiet al., 2016 ] Zhicheng Cui, Wenlin Chen, and Yixin Chen. Multi-scale convolutional neural networks for time series classification.arXiv preprint arXiv:1603.06995,

    2016 arXiv

  33. [2025]

    A time series is worth 64 words: Long-term forecasting with transformers.arXiv preprint arXiv:2211.14730,

    [Nieet al., 2022 ] Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers.arXiv preprint arXiv:2211.14730,

    2022 arXiv

Pith tools

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