Pith. sign in

REVIEW 4 major objections 5 minor 40 references

SEMixer claims that a lightweight MLP-Mixer with random masking and progressive adjacent-scale mixing forecasts long-horizon time series with 5–15% lower error than existing models while using less memory.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 22:36 UTC pith:O4SJ5QGM

load-bearing objection SEMixer is a plausible, efficient multiscale forecasting architecture with strong benchmark numbers, but the RAM mechanism is a global-pooling trick in disguise and the main table has duplicated baselines. the 4 major comments →

arxiv 2602.16220 v3 pith:O4SJ5QGM submitted 2026-02-18 cs.LG

SEMixer: Semantics Enhanced MLP-Mixer for Multiscale Mixing and Long-term Time Series Forecasting

Xu Zhang , Qitong Wang , Peng Wang , Wei Wang This is my paper
classification cs.LG
keywords time series forecastingmultiscale modelingMLP-Mixerrandom attention mechanismdropout ensemblelong-term forecastingpatch-based modelsefficient deep learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper tries to establish that an MLP-Mixer architecture, augmented with two techniques, can be the best practical choice for long-term time series forecasting: it should be more accurate than transformer-, CNN-, and linear-based multiscale models, and also cheaper. The first technique, Random Attention Mechanism (RAM), replaces self-attention with randomly sampled binary interaction masks during training and a dropout-style ensemble at inference to enrich what each time patch represents. The second, Multiscale Progressive Mixing Chain (MPMC), mixes only adjacent temporal scales in a fine-to-coarse chain instead of all scales jointly, which the paper argues reduces memory and avoids semantic gaps between distant scales. If right, this matters because it points to a lower-cost, longer-input route to forecasting accuracy, corroborated by wins on ten public benchmarks and a large real wireless-network operations dataset.

Core claim

The central claim is that a deliberately simple, attention-free MLP-Mixer can outperform 12 advanced forecasting baselines on long-horizon prediction when its patched inputs are semantically enriched by random interaction masks and fused through a progressive adjacent-scale chain. On the paper's own terms, RAM learns a large number of diverse patch-interaction patterns during training and integrates them at inference through a dropout-ensemble approximation, giving MLP-Mixer richer patch semantics than standard self-attention. MPMC then stacks these mixing blocks so each scale is only ever concatenated with its neighbor, moving from finest to coarsest resolution, which the authors say both r

What carries the argument

The two load-bearing pieces are (1) RAM, a random attention mechanism that samples 0–1 Bernoulli masks over a full all-ones interaction matrix during training and, at inference, replaces the ensemble of masks by the all-ones matrix scaled by (1−p) — a dropout-style approximation; and (2) MPMC, a multiscale progressive mixing chain that feeds pairwise concatenations of adjacent scales through RAM+MLP-Mixer blocks in fine-to-coarse order. RAM's job is to give each time patch richer semantics than the raw patch alone, while MPMC's job is to fuse scales without jointly processing all of them.

Load-bearing premise

The semantic-enhancement benefit of RAM rests on the assumption that averaging the learned random masks at inference is faithfully reproduced by a single scaled all-ones matrix — but that matrix collapses to global average pooling, so the 'diverse patch interactions' claimed at inference are not actually being computed.

What would settle it

Run SEMixer with RAM replaced at inference by exact global average pooling (the all-ones matrix scaled by 1−p, without any learned masks) and compare MSE to the full RAM; if accuracy is unchanged, the random-mask training story is falsified and the gain is just global context plus residual.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If SEMixer is right, very long historical inputs can be exploited for forecasting without the memory blow-up that sinks transformer-based multiscale models, because MPMC never processes all scales at once.
  • Standard self-attention's role in patch-based forecasting can be replaced by a cheaper random-mask plus dropout mechanism, with accuracy gains rather than losses, according to the ablation tables.
  • The progressive fine-to-coarse chain is a concrete recipe for aligning scales that differ in semantics, which should make multiscale forecasting more robust to non-adjacent-scale noise.
  • The method's consistency across 10 datasets and a real competition dataset suggests the combination generalizes beyond the specific benchmarks.
  • The noise-injection experiments imply the architecture degrades gracefully when test data are corrupted, with MPMC suppressing noise through shared cross-scale structure.

Where Pith is reading between the lines

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

  • Editorial inference: At inference, RAM's all-ones matrix times a scaled version of the input is mathematically equivalent to global average pooling of the patch columns; so the empirical gains attributed to 'learned diverse interactions' may actually come from adding a global context term plus the residual connection. A direct ablation replacing RAM with exact global pooling would settle this.
  • Editorial inference: The random binary masks during training act as a strong regularizer or denoiser on the interaction matrix; this insight could transfer to other layer types (e.g., convolutions or linear layers) as a general train-time augmentation.
  • Editorial inference: MPMC's adjacent-scale-only mixing is a hierarchical fusion strategy that could be applied to other backbones, not just MLP-Mixer, potentially extending the efficiency and noise-robustness benefits to attention- and convolution-based forecasters.
  • Editorial inference: The paper's own limitation paragraph concedes that RAM assumes meaningful cross-patch correlations; where those correlations are weak, the residual connection must carry the load, suggesting a boundary condition on where SEMixer's gains persist.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes SEMixer, a lightweight multiscale MLP-Mixer architecture for long-term time series forecasting. Two components are introduced: Random Attention Mechanism (RAM), which samples random binary patch-interaction masks during training and aggregates them at inference via a dropout-style ensemble, and Multiscale Progressive Mixing Chain (MPMC), which stacks temporal mixing blocks over pairwise adjacent scales. The model is evaluated on ten public benchmarks, longer-horizon settings, and the 2025 CCF AlOps Challenge, where the authors report third place. The central claims are that RAM enhances patch semantics more effectively and efficiently than self-attention, and that MPMC improves multiscale mixing while reducing memory cost.

Significance. If the results are taken at face value, SEMixer is a simple, efficient, and strong multiscale forecaster. The code is public, ablations cover multiple variants, multiple seeds are reported, and the longer-horizon experiments (Table 2, Table 5) go beyond the standard 96/192/336/720 setup. The MPMC design is a reasonable engineering contribution, and the RAM idea, when correctly characterized, may still be a useful global-context mechanism. However, the current manuscript has several load-bearing reporting and conceptual issues that must be addressed before the claims can be accepted.

major comments (4)
  1. [Table 1 (ILI and Traffic rows)] The baseline entries for ILI and Traffic are duplicated. For ILI, TSMixer, DLinear, PatchTST, and TimeMixer are listed as 2.799/2.758/2.898/2.752, which are identical to Deform.TST/TimeXer/ModernTCN/Pathformer. The correct values in Table 10 give TimeMixer an average ILI MSE of about 2.51, not 2.75. For Traffic, the duplicated TimeMixer value 0.412 is actually Pathformer's; Table 10 shows TimeMixer's true average is about 0.390, which is slightly better than SEMixer's 0.388, reversing the comparison on that dataset. The main results table therefore misrepresents the relative performance on two of ten datasets and directly supports the paper's '5–15% lower MSE' claim. The authors must correct Table 1 using the appendix numbers and re-verify all headline improvements.
  2. [Abstract and Sec. 4.2.2] The paper states in the abstract that SEMixer 'achieves third place' in the 2025 CCF AlOps Challenge, while Sec. 4.2.2 says 'SEMixer attains the lowest prediction error among all competitors' and Table 3 reports the best MSE among the listed methods. If the error is truly the lowest, the model would be first place, not third. The likely resolution is that Table 3 compares only the methods re-run by the authors, not the official leaderboard entries, but this is not stated. This contradiction undermines the competition claim and needs explicit clarification.
  3. [Sec. 3.2, Eq. (5)–(7)] The central mechanism claim for RAM is not supported by the equations. In Eq. (5), the sampled interaction matrix is M ⊙ A, where A is all-ones. Since E[M] = (1-p)J (J all-ones), the dropout ensemble in Eq. (7) is exactly the expected value: the operation becomes (1-p) times the sum over all patches, i.e., a global average pooling per feature. This is identical for every patch; no diversity remains at inference. The claim that RAM 'learns diverse interactions' and integrates them via a dropout ensemble is therefore a mischaracterization of a standard global-context operation. The performance gains attributed to RAM (Table 4) could be explained by global pooling plus residual connections rather than by any learned diversity. The authors should either reformulate RAM so that diverse interaction patterns survive at inference (e.g., learnable non-uniform weights, multiple masks, or per-head
  4. [Sec. 4.2.1 and Table 1] The claim 'SEMixer achieves superior long-term forecasting performance, with 5–15% lower MSE than existing models' is an overgeneralization. Using the correct Table 1 entries, several datasets show improvements well below 5% (e.g., Solar Energy ~0%, ETTh1 ~2%, Electricity ~1%), and with the duplicated entries corrected, Traffic shows no improvement over TimeMixer. The percentage range should be recomputed on corrected numbers and stated per dataset, or the claim should be removed.
minor comments (5)
  1. [Sec. 3.2] The 'Training stage of RAM' paragraph is repeated nearly verbatim twice. Please remove the duplicate.
  2. [Sec. 3.1, Eq. (3)] The patch count formula N_s = floor((n-L_s)/K_s)+2 should be justified. The standard formula is +1 for non-overlapping patches; the +2 appears to imply an extra padding or a different convention. Please clarify.
  3. [Sec. 4.1.3 and Figure 1 caption] It is unclear whether the optimal input length was also searched for SEMixer or only for the baselines. The text says 'we also search for the optimal input length ... for each baseline', and the Figure 1 caption says SEMixer adjusts only input length. A precise statement of the selection protocol for SEMixer and each baseline is needed to guarantee fairness.
  4. [Sec. 4.2.2, Table 3] The table reports only six methods; the list of 'advanced methods' compared is incomplete relative to the 12 baselines in Table 1. State why FiLM, Scaleformer, and others are omitted from the competition comparison.
  5. [General] There are several typos and formatting inconsistencies, e.g., 'Mutiscale' in Sec. 3.1, the use of 'bX_d' in Sec. 4.1.3 where a dimension name is expected, and the incomplete 'w/ FA' column in Table 4 (missing values for some rows).

Circularity Check

0 steps flagged

No significant circularity: central claims rest on external benchmarks; RAM's Eq. 7 being global pooling is a mechanism critique, not a circular derivation.

full rationale

The paper's prediction claims (MSE improvements on 10 public datasets and the 2025 CCF AlOps Challenge) are evaluated against external benchmarks and an official competition platform; no parameter is fitted to the target results and then reported as a prediction. The model's hyperparameters (e.g., p=0.85, scale factors) are fixed choices, not derived from benchmark outcomes. The introduction cites the authors' prior works (Refs 33-35) for motivation such as financial sequences and historical-data forecasting, but these citations are not load-bearing: none is invoked as a uniqueness theorem or as justification for SEMixer's architecture. The only notable issue is that Eqs. 5-7 show the inference-time RAM operation is the expectation over Bernoulli masks, which equals (1-p) times an all-ones matrix acting on X — i.e., scaled global pooling — so the paper's rhetorical claim that RAM 'learns diverse interactions' that persist at inference is mechanistically unsupported. However, this is a novelty/mechanism critique, not a circular derivation: the empirical gains are measured, not derived by construction from the definition of RAM. Accordingly, no circularity step meets the evidentiary bar.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The central claim depends on hand-chosen hyperparameters (p, S, scale factors, patch count, hidden dims) and on several domain assumptions about patching, MLP-Mixer's expressiveness, and the existence of 'semantic gaps.' No new physical entities are introduced. The most fragile assumption is that RAM's inference-time global pooling (Eq. 7) constitutes learned semantic enhancement rather than a generic global-context bias.

free parameters (5)
  • sampling disconnection probability p = 0.85
    Chosen by hand in Sec. 4.1.3; controls the Bernoulli mask in RAM and the (1-p) scaling at inference. Its value directly sets the strength of the global-pooling term.
  • number of scales S = 4
    Set to 4 in Sec. 4.1.3; determines how many multiscale patch resolutions are mixed in MPMC.
  • scale factors alpha_2, alpha_3, alpha_4 = 2, 4, 8
    Fixed in Sec. 4.1.3; define the patch-length and stride ratios across the four scales.
  • finest-scale patch count N1 = 64
    Fixed in Sec. 4.1.3; sets the number of patches at the finest scale, which anchors all other scale patch counts.
  • embedding and hidden dimensions D and bX_d = 128 and 64
    Fixed in Sec. 4.1.3; standard capacity choices, not tuned per dataset.
axioms (4)
  • domain assumption Time series can be faithfully represented by overlapping patches at multiple scales
    Used throughout Sec. 3.1; if patching destroys phase or level information, the multiscale inputs are lossy and the architecture cannot recover it.
  • standard math MLP-Mixer (inter/intra-patch MLPs) can capture temporal dependencies without attention
    Borrowed from TSMixer/MLP-Mixer [24,6]; not proven in this paper, but an established empirical result in the cited literature.
  • standard math Expectation of random masks over training approximates the ensemble at inference (dropout ensemble principle)
    Used in Eq. 7; well-known from Dropout [23] and prior ensemble analyses.
  • ad hoc to paper Semantic gaps exist between non-adjacent scales and pairwise adjacent-scale mixing mitigates them
    Qualitative claim motivating MPMC (Sec. 1 and Sec. 3.3); no quantitative evidence is provided beyond ablation, and the concept is not formally defined.

pith-pipeline@v1.3.0-alltime-deepseek · 25245 in / 14527 out tokens · 134609 ms · 2026-08-02T22:36:24.376112+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of SEMixer: Semantics Enhanced MLP-Mixer for Multiscale Mixing and Long-term Time Series Forecasting." pith.science (2026). https://pith.science/paper/O4SJ5QGM

@misc{pith2026260216220,
  author       = {Pith},
  title        = {Pith review of: SEMixer: Semantics Enhanced MLP-Mixer for Multiscale Mixing and Long-term Time Series Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/O4SJ5QGM}},
  note         = {Machine review of arXiv:2602.16220}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Modeling multiscale patterns is crucial for long-term time series forecasting (TSF). However, redundancy and noise in time series, together with semantic gaps between non-adjacent scales, make the efficient alignment and integration of multi-scale temporal dependencies challenging. To address this, we propose SEMixer, a lightweight multiscale model designed for long-term TSF. SEMixer features two key components: a Random Attention Mechanism (RAM) and a Multiscale Progressive Mixing Chain (MPMC). RAM captures diverse time-patch interactions during training and aggregates them via dropout ensemble at inference, enhancing patch-level semantics and enabling MLP-Mixer to better model multi-scale dependencies. MPMC further stacks RAM and MLP-Mixer in a memory-efficient manner, achieving more effective temporal mixing. It addresses semantic gaps across scales and facilitates better multiscale modeling and forecasting performance. We not only validate the effectiveness of SEMixer on 10 public datasets, but also on the \textit{2025 CCF AlOps Challenge} based on 21GB real wireless network data, where SEMixer achieves third place. The code is available at the link https://github.com/Meteor-Stars/SEMixer.

Figures

Figures reproduced from arXiv: 2602.16220 by Peng Wang, Qitong Wang, Wei Wang, Xu Zhang.

Figure 1
Figure 1. Figure 1: Sub-figures (a)-(e): evaluate whether popular multi- and single-scale models benefit from longer historical sequences, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: SEMixer components: The Multiscale Encoding Block processes historical input [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visualizing learned 12000 patch embeddings of 4 scale inputs [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Further ablation study of MPMC structure. Average [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Hyperparameter analysis of the random disconnection probability [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

40 extracted references · 1 canonical work pages

  1. [1]

    Sadowski

    Pierre Baldi and Peter J. Sadowski. 2013. Understanding Dropout. InAdvances in Neural Information Processing Systems 26: 27th Annual Conference on Neural Infor- mation Processing Systems 2013. Proceedings of a meeting held December 5-8, 2013, Lake Tahoe, Nevada, United States, Christopher J. C. Burges, Léon Bottou, Zoubin Ghahramani, and Kilian Q. Weinber...

    2013

  2. [2]

    Peng Chen, Yingying Zhang, Yunyao Cheng, Yang Shu, Yihang Wang, Qing- song Wen, Bin Yang, and Chenjuan Guo. 2024. Pathformer: Multi-scale trans- formers with Adaptive Pathways for Time Series Forecasting.arXiv preprint arXiv:2402.05956(2024)

    Pith/arXiv arXiv 2024

  3. [3]

    Si-An Chen, Chun-Liang Li, Nate Yoder, Sercan O Arik, and Tomas Pfister. 2023. Tsmixer: An all-mlp architecture for time series forecasting.arXiv preprint arXiv:2303.06053(2023)

    Pith/arXiv arXiv 2023

  4. [4]

    Krzysztof Marcin Choromanski, Valerii Likhosherstov, David Dohan, Xingyou Song, Andreea Gane, Tamas Sarlos, Peter Hawkins, Jared Quincy Davis, Afroz Mohiuddin, Lukasz Kaiser, et al. [n. d.]. Rethinking Attention with Performers. InInternational Conference on Learning Representations

  5. [5]

    Francis X Diebold and Georg Strasser. 2013. On the correlation structure of microstructure noise: A financial economic approach.Review of Economic Studies 80, 4 (2013), 1304–1337

    2013

  6. [6]

    Vijay Ekambaram, Arindam Jati, Nam Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. 2023. TSMixer: Lightweight MLP-Mixer Model for Multivariate Time Series Forecasting. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2023, Long Beach, CA, USA, August 6-10, 2023, Ambuj K. Singh, Yizhou Sun, Leman Akoglu, Dimitr...

    arXiv 2023

  7. [7]

    Kazuyuki Hara, Daisuke Saitoh, and Hayaru Shouno. 2016. Analysis of Dropout Learning Regarded as Ensemble Learning. InArtificial Neural Networks and Machine Learning - ICANN 2016 - 25th International Conference on Artificial Neural Networks, Barcelona, Spain, September 6-9, 2016, Proceedings, Part II (Lecture Notes in Computer Science, Vol. 9887), Alessan...

    work page doi:10.1007/978-3-319-44781-0_9 2016

  8. [8]

    Min Hou, Chang Xu, Zhi Li, Yang Liu, Weiqing Liu, Enhong Chen, and Jiang Bian. 2022. Multi-Granularity Residual Learning with Confidence Estimation for Time Series Prediction. InProceedings of the ACM Web Conference 2022. 112–121

    2022

  9. [9]

    Xinrui Jiang, Yicheng Pan, Meng Ma, and Ping Wang. 2023. Look Deep into the Microservice System Anomaly through Very Sparse Logs. InProceedings of the ACM Web Conference 2023. 2970–2978

    2023

  10. [10]

    D Kim, J Park, J Lee, and H Kim. 2024. Are Self-Attentions Effective for Time Series Forecasting?. In38th Conference on Neural Information Processing Systems (NeurIPS 2024)

    2024

  11. [11]

    Taesung Kim, Jinhee Kim, Yunwon Tae, Cheonbok Park, Jang-Ho Choi, and Jaegul Choo. 2022. Reversible Instance Normalization for Accurate Time-Series Forecasting against Distribution Shift. InThe Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenRe- view.net. https://openreview.net/forum?id=cGDAkQo1C0p

    2022

  12. [12]

    Nikita Kitaev, Lukasz Kaiser, and Anselm Levskaya. [n. d.]. Reformer: The Efficient Transformer. InInternational Conference on Learning Representations

  13. [13]

    Shiyang Li, Xiaoyong Jin, Yao Xuan, Xiyou Zhou, Wenhu Chen, Yu-Xiang Wang, and Xifeng Yan. 2019. Enhancing the locality and breaking the memory bottle- neck of transformer on time series forecasting.Advances in neural information processing systems32 (2019)

    2019

  14. [14]

    Shizhan Liu, Hang Yu, Cong Liao, Jianguo Li, Weiyao Lin, Alex X Liu, and Schahram Dustdar. 2021. Pyraformer: Low-complexity pyramidal attention for long-range time series modeling and forecasting. InInternational conference on learning representations

    2021

  15. [15]

    Xu Liu, Junfeng Hu, Yuan Li, Shizhe Diao, Yuxuan Liang, Bryan Hooi, and Roger Zimmermann. 2024. Unitime: A language-empowered unified model for cross- domain time series forecasting. InProceedings of the ACM on Web Conference

    2024

  16. [16]

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. 2024. iTransformer: Inverted Transformers Are Effective for Time Series Forecasting. InThe Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net. https://openreview.net/forum?id=JePfAI8fah

    2024

  17. [17]

    Donghao Luo and Xue Wang. 2024. DeformableTST: Transformer for Time Series Forecasting without Over-reliance on Patching. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems

    2024

  18. [18]

    Donghao Luo and Xue Wang. 2024. Moderntcn: A modern pure convolution structure for general time series analysis. InThe twelfth international conference on learning representations. 1–43

    2024

  19. [19]

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

    Pith/arXiv arXiv 2022

  20. [20]

    P Nunes, J Santos, and E Rocha. 2023. Challenges in predictive maintenance–A review.CIRP Journal of Manufacturing Science and Technology40 (2023), 53–67

    2023

  21. [21]

    Diptangshu Pandit, Li Zhang, Chengyu Liu, Nauman Aslam, Samiran Chattopad- hyay, and Chee Peng Lim. 2017. Noise reduction in ECG signals using wavelet transform and dynamic thresholding.Emerging trends in neuro engineering and neural computation(2017), 193–206

    2017

  22. [22]

    Amin Shabani, Amir Abdi, Lili Meng, and Tristan Sylvain. 2022. Scaleformer: iterative multi-scale refining transformers for time series forecasting.arXiv preprint arXiv:2206.04038(2022)

    Pith/arXiv arXiv 2022

  23. [23]

    Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. Dropout: a simple way to prevent neural networks from overfitting.The journal of machine learning research15, 1 (2014), 1929–1958

    2014

  24. [24]

    Ilya O Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Andreas Steiner, Daniel Keysers, Jakob Uszkoreit, et al. 2021. Mlp-mixer: An all-mlp architecture for vision.Advances in neural information processing systems34 (2021), 24261–24272

    2021

  25. [25]

    Laurens Van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. Journal of machine learning research9, 11 (2008)

    2008

  26. [26]

    Shiyu Wang, Jiawei Li, Xiaoming Shi, Zhou Ye, Baichuan Mo, Wenze Lin, Sheng- tong Ju, Zhixuan Chu, and Ming Jin. 2024. Timemixer++: A general time series pattern machine for universal predictive analysis.arXiv preprint arXiv:2410.16032 (2024)

    Pith/arXiv arXiv 2024

  27. [27]

    Shiyu Wang, Haixu Wu, Xiaoming Shi, Tengge Hu, Huakun Luo, Lintao Ma, James Y Zhang, and JUN ZHOU. 2024. TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting. InThe Twelfth International Conference on Learning Representations

    2024

  28. [28]

    Yuxuan Wang, Haixu Wu, Jiaxiang Dong, Guo Qin, Haoran Zhang, Yong Liu, Yunzhong Qiu, Jianmin Wang, and Mingsheng Long. [n. d.]. TimeXer: Empower- ing Transformers for Time Series Forecasting with Exogenous Variables. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems

  29. [29]

    Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. [n. d.]. TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis. InThe Eleventh International Conference on Learning Representations

  30. [30]

    Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. 2021. Autoformer: De- composition transformers with auto-correlation for long-term series forecasting. Advances in Neural Information Processing Systems34 (2021), 22419–22430

    2021

  31. [31]

    Wentao Xu, Weiqing Liu, Chang Xu, Jiang Bian, Jian Yin, and Tie-Yan Liu. 2021. Rest: Relational event-driven stock trend forecasting. InProceedings of the web conference 2021. 1–10

    2021

  32. [32]

    Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu. 2023. Are transformers effective for time series forecasting?. InProceedings of the AAAI conference on artificial intelligence, Vol. 37. 11121–11128

    2023

  33. [33]

    Xu Zhang, Zhengang Huang, Yunzhi Wu, Xun Lu, Erpeng Qi, Yunkai Chen, Zhongya Xue, Peng Wang, and Wei Wang. 2024. Self-Adaptive Scale Handling for Forecasting Time Series with Scale Heterogeneity. InICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 7485–7489

    2024

  34. [34]

    Xu Zhang, Zhengang Huang, Yunzhi Wu, Xun Lu, Erpeng Qi, Yunkai Chen, Zhongya Xue, Qitong Wang, Peng Wang, and Wei Wang. 2025. Multi-period learning for financial time series forecasting. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 1. 2848–2859

    2025

  35. [35]

    Xu Zhang, Qitong Wang, Peng Wang, and Wei Wang. 2025. A Lightweight Sparse Interaction Network for Time Series Forecasting. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 13304–13312

    2025

  36. [36]

    Yunhao Zhang and Junchi Yan. 2023. Crossformer: Transformer utilizing cross- dimension dependency for multivariate time series forecasting. InThe eleventh international conference on learning representations

    2023

  37. [37]

    Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wancai Zhang. 2021. Informer: Beyond efficient transformer for long se- quence time-series forecasting. InProceedings of the AAAI conference on artificial intelligence, Vol. 35. 11106–11115

    2021

  38. [38]

    Tian Zhou, Ziqing Ma, Qingsong Wen, Liang Sun, Tao Yao, Wotao Yin, Rong Jin, et al. 2022. Film: Frequency improved legendre memory model for long-term time series forecasting.Advances in Neural Information Processing Systems35 (2022), 12677–12690

    2022

  39. [39]

    Tian Zhou, Ziqing Ma, Qingsong Wen, Xue Wang, Liang Sun, and Rong Jin

  40. [2022]

    InInternational Conference on Machine Learning

    Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting. InInternational Conference on Machine Learning. PMLR, 27268– 27286. A Appendices A.1 Datasets The 10 public datasets used in this paper are extensively used for long-term TSF algorithm evaluation, covering multiple fields in- cluding industry (4 ETT datasets), climate (...

    2026