Pith. sign in

REVIEW 4 major objections 6 minor 41 references

SPAT: Sensitivity-based Multihead-attention Pruning on Time Series Forecasting Models

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read SPAT removes redundant multi-head attention modules from transformer forecasters, cutting FLOPs and improving forecast accuracy.

desk verdict A useful pruning recipe with real speedups, but the headline mechanism—SEND—is not actually tested in the configuration that drives most of the gains. read the letter →

arxiv 2505.08768 v1 pith:5WZBHVDQ submitted 2025-05-13 cs.LG

classification cs.LG
keywords attentionpruningstructuredtimeseriesforecastingmulti-headsensitivityanalysismodelcompressionzero-shottransfertransformer
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

SPAT is a structured pruning method for multivariate time series forecasting. Its central claim is that many whole multi-head attention (MHA) modules inside transformer forecasters are redundant or harmful, and that deleting them—not pruning individual weights or heads—makes the model smaller and faster while slightly improving forecast accuracy. The paper proposes a dynamic sensitivity metric, SEND, which ranks attention layers by the dispersion of their gradient-based importance scores, removes the lowest-ranked layers, and fine-tunes the surviving network. Across PatchTST and iTransformer on eight datasets, the pruned models achieve average reductions of 2.842% in MSE, 1.996% in MAE, and 35.274% in FLOPs, and they beat lightweight, Mamba-based, and LLM-based baselines in both standard and zero-shot inference. The point of interest is that expensive attention machinery can be cut wholesale, without specialized hardware, and still preserve or improve generalization.

What carries the argument

The central object is the SEND (Sensitivity Enhanced Normalized Dispersion) score, computed per attention layer. For each layer, SPAT builds a sensitivity matrix as the gradient of the average training loss with respect to a binary connection mask over the attention score matrix, takes absolute values, normalizes each row with a softmax, averages across attention heads, and then averages the row-wise standard deviations of the resulting matrix. That scalar ranks the layer; the bottom-K layers are removed and the model is fine-tuned. The load-bearing mechanism is the assumed link between dispersion of normalized gradient importance and the usefulness of an attention layer, as opposed to degenerate attention that behaves like a scaled identity mapping.

What would settle it

A reader could take a transformer forecaster, compute SEND rankings, and compare fine-tuned performance after removing the lowest-ranked layer against removing a randomly chosen layer over several seeds; if random removal matches SEND removal on average, the metric is not doing the ranking work.

Watch

Extended reading notes

Core claim

The paper claims that transformer-based forecasters are over-provisioned with attention: many MHA layers contribute little or actively hurt performance, and removing whole layers chosen by a sensitivity-based importance score improves efficiency and accuracy. The discovery is that a single dispersion statistic, SEND, can identify those layers during pretraining. SEND is the average row-wise standard deviation of a softmax-normalized sensitivity matrix, where sensitivity is the gradient of the loss with respect to a binary connection mask on the attention scores. On the paper's experiments, pruned PatchTST improves average MSE by 1.266% and MAE by 0.824% with 13.990% fewer FLOPs, while pruned iTransformer improves MSE by 4.418% and MAE by 3.167% with over 50% fewer FLOPs. The pruned models also retain zero-shot forecasting ability across datasets, outperforming the compared lightweight, Mamba-based, and LLM-based baselines.

Load-bearing premise

The load-bearing premise is that a low SEND score—low dispersion in the normalized sensitivity gradients—truly flags an attention layer as redundant, so removing the lowest-scoring layers is better than removing random layers.

Editorial extensions

If this is right

  • Transformer forecasters contain whole attention layers that can be removed without hurting accuracy, and on most datasets removal improves forecast error.
  • Module-level pruning gives speedups on ordinary GPUs because the surviving model is a smaller transformer, unlike unstructured weight pruning that needs specialized hardware.
  • Keeping the most effective attention layers is what sustains zero-shot forecasting across datasets; pruning too aggressively ($\alpha=0.9$ on PatchTST) degrades performance.
  • The optimal number of layers to remove depends on the architecture: PatchTST's best result comes from removing one layer, while iTransformer improves even after removing several.
  • SPAT sits on top of existing transformer forecasters unchanged, so it can be combined with patching, masking, or other efficiency techniques.

Reading between the lines

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

  • Because SEND uses only attention scores and gradients, the same ranking could be tested on attention layers in vision transformers or language models, where whole-MHA pruning has been explored with other criteria.
  • A direct way to isolate the metric's contribution is to compare SEND-ranked removal against random layer removal with the same fine-tuning budget; if random matches SEND, the gains come from fine-tuning and architecture, not from the score.
  • iTransformer's attention operates over channels rather than time patches, and it benefits more from pruning; this suggests channel-dimension attention is more redundant, so SPAT's gains may grow with the number of series channels.
  • The pretraining-and-finetuning loop leaves room to stack other compression techniques, such as head pruning or quantization on surviving layers, so the reported FLOP reductions are not necessarily the ceiling.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes SPAT, a structured pruning method that removes entire multi-head attention (MHA) modules from transformer-based time series forecasting models. The importance of each MHA module is measured during a pretraining phase by SEND (Sensitivity Enhanced Normalized Dispersion), computed as the average row-wise standard deviation of softmax-normalized sensitivity gradients. Layers with the lowest SEND scores are removed according to a pruning ratio α, and the model is fine-tuned. Experiments on eight multivariate datasets with PatchTST and iTransformer report average reductions of 2.842% in MSE, 1.996% in MAE, 35.274% in FLOPs, and 28.191% in parameters, as well as claims that the pruned models outperform lightweight, Mamba-based, and LLM-based baselines in both standard and zero-shot inference. The paper also includes ablation studies on the pruning ratio, lookback window length, model efficiency, and a limitations section.

Significance. If the results hold, the paper's central idea is valuable: pruning entire attention modules, rather than individual weights or heads, could make transformer forecasters smaller and faster without specialized hardware, and the reported accuracy gains over the original models are surprising and interesting. The SEND metric itself is simple, contains no fitted constants, and the authors provide code and detailed per-horizon results in the appendix, which strengthens reproducibility. The paper also honestly lists limitations in Appendix D. However, the empirical support for the specific claim that SEND-based selection, rather than the act of removing attention layers plus fine-tuning, drives the improvements is incomplete. The current experiments do not isolate the selection mechanism, and the largest contributor to the headline gains uses a configuration in which the SEND ranking is never exercised. Because the paper's central contribution is the selective pruning criterion, this gap is load-bearing.

major comments (4)
  1. [§4.2, Table 4, Eq. (14)] The pruning ratio α is selected from the test-set results reported in Table 4 and then used to produce the headline improvements in Tables 1–3. This is test-set-based model selection, which makes the reported gains optimistic and undercuts the claim that SPAT improves over the original models in a well-defined way. The manuscript should specify whether α was chosen on a validation split, or should provide a separate evaluation protocol that does not use test data for selecting α.
  2. [§3.3, §4.2, Table 4, Table 6, Eq. (14)] For iTransformer, the best configuration in Table 4 is α=0.9 on all reported datasets, and Table 6 shows that iTransformer has N=2, 3, or 4 attention layers. With K=ceil(αN), α=0.9 removes all MHA modules (K=N) for every iTransformer configuration. Therefore the SEND ranking in Eq. (13) is never used for iTransformer; the pruned iTransformer results in Tables 1–3 are simply the original model with every attention layer removed and then fine-tuned. Since iTransformer contributes an average MSE improvement of 4.418%, more than two-thirds of the combined 2.842% headline reduction comes from a configuration that bypasses the proposed selection mechanism. The authors should either report results for a configuration where SEND actually selects a subset of layers, or explicitly separate the claim 'removing all attention layers helps iTransformer' from the claim 'SEND selects which layers to remove.'
  3. [§4.4, Table 4] No experiment isolates the contribution of the SEND ranking. For PatchTST, α=0.3 removes one of three attention layers, so SEND does choose a layer, but the ablation only varies α and never compares the SEND-selected layer against a random layer, the first/last layer, or the layer with the lowest/highest SEND under the same fine-tuning protocol. Without such a baseline, the observed improvements could result from capacity reduction and fine-tuning rather than from the proposed importance metric. The authors should add a fixed-layer or random-layer pruning baseline and report the identity of the pruned layer selected by SEND.
  4. [§4.2, Table 3, Abstract] The abstract and contributions state that SPAT-pruned models outperform lightweight, Mamba-based, and LLM-based methods in both standard and zero-shot inference, but Table 3 does not support this for pruned iTransformer. For example, in ETTh1→ETTh2 pruned iTransformer has MSE 0.363 versus TSLANet 0.359; in ETTh2→ETTh1 pruned iTransformer has MSE 0.596 versus the best 0.496; and pruned iTransformer is also worse than several baselines in ETTm1→ETTh2, ETTm2→ETTh1, and ETTm2→ETTh2. The zero-shot claim should be restricted to pruned PatchTST or otherwise qualified according to the actual table results.
minor comments (6)
  1. [Table 1] There are formatting errors in Table 1, such as "0 .43" in the ETTh1 row and irregular spacing in several cells; these should be cleaned up for readability.
  2. [Table 4] Table 4 omits the ETTh2 and ETTm2 datasets that appear in the other main tables. The authors should either include them or explain why they are excluded from the pruning-ratio ablation.
  3. [References [7] and [15]] References [7] (Approximate attention with MLP) and [15] (Bi-Mamba+) both list the same arXiv identifier 2404.15772, which appears to be a duplicate or a citation error. Please correct the reference for Bi-Mamba+.
  4. [Table 3] The table uses the symbols ⋆ and † in the caption but the corresponding footnotes are not present in the text; please either add the footnotes or remove the symbols.
  5. [Algorithm 1] Algorithm 1 introduces a state set S that is not further explained in the text; the notation should be defined or removed.
  6. [§4.4] The sentence "For PatchTST, α = 0.9, which removes all MHA modules, leads to a performance drop" is correct only because PatchTST has N=3 layers and ceil(0.9*3)=3; the same logic makes α=0.9 vacuous for iTransformer with N≤4, which is directly relevant to the major comment above and should at least be acknowledged in the ablation discussion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; SEND is an independently defined gradient-dispersion metric, and the reported gains are empirical results rather than outputs forced by the metric's definition.

full rationale

The derivation chain is self-contained. SEND is computed from raw gradients (Eqs. 8–12, 16–19) and contains no fitted constants; it is not constructed from the MSE/MAE improvements it is later compared against. The pruning procedure applies SEND to rank layers, and the reported gains are empirical outcomes of pruning plus fine-tuning, so they are not equal to SEND by construction. The choice of pruning ratio α in Table 4 is tuned on the same benchmarks, and for iTransformer α=0.9 removes all MHA modules, so the SEND ranking is not exercised for that model; however, this is an attribution/evaluation confound, not a circular reduction in which a fitted parameter is renamed as a prediction. The self-citations ([6], [7]) are motivational (degenerate attention as identity-like, sensitivity pruning inspiration) and are not load-bearing for the central derivation, which is supported by the paper's own experiments. No quoted equation makes the claimed improvement depend on the metric by definition.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim rests on a small set of design choices: the gradient-as-sensitivity approximation, the dispersion-as-effectiveness premise, and the per-architecture pruning ratio alpha. No new physical entities are introduced, but the SEND metric itself is a new heuristic with no external falsifiable handle beyond the reported forecasting benchmarks.

free parameters (1)
  • pruning ratio alpha = 0.3 for PatchTST, 0.9 for iTransformer
    Alpha is chosen per architecture after inspecting test-set performance in Table 4, then used for the main reported results. This is a data-dependent hyperparameter selection that affects the headline gains.
assumptions (3)
  • domain assumption First-order sensitivity approximation: the discrete loss change from removing an attention connection is approximated by the gradient of the loss with respect to a relaxed mask at mask=1.
    The paper uses this in Eq. (8) to define sensitivity. Neural-network losses are not linear in the mask, so the first-order Taylor approximation is a standard heuristic, not an exact identity.
  • ad hoc to paper Higher dispersion of the SEND sensitivity matrix indicates higher effectiveness of the attention layer.
    This is stated in Section 3.2 after Eq. (19) and is the load-bearing conceptual premise for ranking and pruning. It is asserted from Figure 1 rather than derived.
  • ad hoc to paper Row-wise softmax normalization of absolute sensitivities preserves relative importance across attention positions and layers.
    Introduced in Eq. (11). The choice of softmax with no temperature is a design decision without independent validation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SPAT: Sensitivity-based Multihead-attention Pruning on Time Series Forecasting Models." pith.science (2026). https://pith.science/paper/5WZBHVDQ

@misc{pith2026250508768,
  author       = {Pith},
  title        = {Pith review of: SPAT: Sensitivity-based Multihead-attention Pruning on Time Series Forecasting Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5WZBHVDQ}},
  note         = {Machine review of arXiv:2505.08768}
}
abstract

Attention-based architectures have achieved superior performance in multivariate time series forecasting but are computationally expensive. Techniques such as patching and adaptive masking have been developed to reduce their sizes and latencies. In this work, we propose a structured pruning method, SPAT ($\textbf{S}$ensitivity $\textbf{P}$runer for $\textbf{At}$tention), which selectively removes redundant attention mechanisms and yields highly effective models. Different from previous approaches, SPAT aims to remove the entire attention module, which reduces the risk of overfitting and enables speed-up without demanding specialized hardware. We propose a dynamic sensitivity metric, $\textbf{S}$ensitivity $\textbf{E}$nhanced $\textbf{N}$ormalized $\textbf{D}$ispersion (SEND) that measures the importance of each attention module during the pre-training phase. Experiments on multivariate datasets demonstrate that SPAT-pruned models achieve reductions of 2.842% in MSE, 1.996% in MAE, and 35.274% in FLOPs. Furthermore, SPAT-pruned models outperform existing lightweight, Mamba-based and LLM-based SOTA methods in both standard and zero-shot inference, highlighting the importance of retaining only the most effective attention mechanisms. We have made our code publicly available https://anonymous.4open.science/r/SPAT-6042.

Figures

Figures reproduced from arXiv: 2505.08768 by the authors.

Figure 1
Figure 1. Attention score visualization: (a) Degenerate vs. (b) Effective. Degenerate attention is [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of attention-based models (a) and pruning algorithm workflow (b). Solid arrows [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Model efficiency comparison with average results over 4 forecasting horizons on Traffic. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Influence of lookback window lengths L 9 [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: The MultiHead attention mechanism structure. [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 34 canonical work pages

  1. [1]

    S. Deng, S. Wang, H. Rangwala, L. Wang, and Y . Ning. Cola-GNN: Cross-location attention based graph neural networks for long-term ILI prediction. In CIKM, pages 245–254, 2020

  2. [2]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In ICLR, 2021

  3. [3]

    W. Duan, X. He, Z. Zhou, L. Thiele, and H. Rao. Localised adaptive spatial-temporal graph neural network. In KDD, pages 448–458, 2023

  4. [4]

    Ekambaram, A

    V . Ekambaram, A. Jati, N. Nguyen, P. Sinthong, and J. Kalagnanam. TSMixer: Lightweight MLP-mixer model for multivariate time series forecasting. In KDD, pages 459–469, 2023

  5. [5]

    Eldele, M

    E. Eldele, M. Ragab, Z. Chen, M. Wu, and X. Li. TSLANet: Rethinking transformers for time series representation learning. In ICML, 2024

  6. [6]

    S. Guo, B. Lai, S. Yang, J. Zhao, and F. Shen. Sensitivity pruner: Filter-Level compression algorithm for deep neural networks. Pattern Recognit., 140:109508, 2023

  7. [8]

    Jiang, C

    J. Jiang, C. Han, W. X. Zhao, and J. Wang. Pdformer: Propagation delay-aware dynamic long-range transformer for traffic flow prediction. In AAAI, pages 4365–4373, 2023

  8. [9]

    M. Jin, S. Wang, L. Ma, Z. Chu, J. Y . Zhang, X. Shi, P.-Y . Chen, Y . Liang, Y .-F. Li, S. Pan, and Q. Wen. Time-LLM: Time series forecasting by reprogramming large language models. In ICLR, 2024

Show all 41 references
  1. [10]

    Kitaev, L

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

  2. [11]

    P. W. Koh and P. Liang. Understanding black-box predictions via influence functions. In ICML, pages 1885–1894, 2017

  3. [12]

    N. Lee, T. Ajanthan, and P. H. Torr. SNIP: Single-shot network pruning based on connection sensitivity. In ICLR, 2019. 10

  4. [13]

    A. W. Li and G. S. Bastos. Stock Market Forecasting Using Deep Learning and Technical Analysis: A Systematic Review. IEEE Access, 8:185232–185242, 2020

  5. [14]

    S. Li, X. Jin, Y . Xuan, X. Zhou, W. Chen, Y .-X. Wang, and X. Yan. Enhancing the locality and breaking the memory bottleneck of transformer on time series forecasting. In NeurIPS, pages 5244–5254, 2019

  6. [15]

    Liang, X

    A. Liang, X. Jiang, Y . Sun, and C. Lu. Bi-Mamba+: Bidirectional Mamba for Time Series Forecasting. CoRR, abs/2404.15772, 2024

  7. [16]

    Liang, S

    C. Liang, S. Zuo, M. Chen, H. Jiang, X. Liu, P. He, T. Zhao, and W. Chen. Super tickets in pre-trained language models: From model compression to improving generalization. In ACL, pages 6524–6538, 2021

  8. [17]

    S. Lin, P. Lyu, D. Liu, T. Tang, X. Liang, A. Song, and X. Chang. MLP can be a good transformer learner. In CVPR, pages 19489–19498, 2024

  9. [18]

    H. Liu, Z. Dong, R. Jiang, J. Deng, J. Deng, Q. Chen, and X. Song. STAEformer: Spatio- temporal adaptive embedding makes vanilla transformer SOTA for traffic forecasting. In CIKM, 2023

  10. [19]

    M. Liu, A. Zeng, M. Chen, Z. Xu, Q. Lai, L. Ma, and Q. Xu. SCINet: Time series modeling and forecasting with sample convolution and interaction. In NeurIPS, 2022

  11. [20]

    S. Liu, H. Yu, C. Liao, J. Li, W. Lin, A. X. Liu, and S. Dustdar. Pyraformer: Low-complexity pyramidal attention for long-range time series modeling and forecasting. In ICLR, 2022

  12. [21]

    Lu, X.-Y

    H. Lu, X.-Y . Chen, H.-J. Ye, and D.-C. Zhan. SOFTS: Efficient multivariate time series forecasting with series-core fusion. In NeurIPS, 2024

  13. [22]

    Michel, O

    P. Michel, O. Levy, and G. Neubig. Are sixteen heads really better than one? In NeurIPS, pages 14014–14024, 2019

  14. [23]

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

  15. [24]

    Sarmas, N

    E. Sarmas, N. Dimitropoulos, V . Marinakis, Z. Mylona, and H. Doukas. Transfer learning strategies for solar power forecasting under data scarcity. Sci Rep, 12(1):14643, 2022

  16. [25]

    Touvron, L

    H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y . Babaei, N. Bashlykov, S. Ba- tra, P. Bhargava, S. Bhosale, D. Bikel, L. Blecher, C. Canton-Ferrer, M. Chen, G. Cucurull, D. Esiobu, J. Fernandes, J. Fu, W. Fu, B. Fuller, C. Gao, V . Goswami, N. Goyal, A. Hartshorn,...

  17. [26]

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

  18. [27]

    S. Wang, B. Z. Li, M. Khabsa, H. Fang, and H. Ma. Linformer: Self-attention with linear complexity. CoRR, abs/2006.04768, 2020

  19. [28]

    H. Wu, J. Xu, J. Wang, and M. Long. Autoformer: Decomposition transformers with auto- correlation for long-term series forecasting. In NeurIPS, volume 34, pages 22419–22430, 2021

  20. [29]

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

  21. [30]

    Z. Wu, S. Pan, G. Long, J. Jiang, X. Chang, and C. Zhang. Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks. In KDD, pages 753–763, 2020. 11

  22. [31]

    M. Xia, Z. Zhong, and D. Chen. Structured pruning learns compact and accurate models. In ACL, pages 1513–1528, 2022

  23. [32]

    X. Xu, Y . Liang, B. Huang, Z. Lan, and K. Shu. Integrating mamba and transformer for long-short range time series forecasting. CoRR, abs/2404.14757, 2024

  24. [33]

    Z. Yang, Y . Cui, X. Yao, and S. Wang. Gradient-based intra-attention pruning on pre-trained language models. In ACL, pages 2775–2790, 2023

  25. [34]

    A. Zeng, M. Chen, L. Zhang, and Q. Xu. Are transformers effective for time series forecasting? In AAAI, pages 11121–11128, 2023

  26. [35]

    H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, and W. Zhang. Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting. In AAAI, pages 11106–11115, 2021

  27. [36]

    T. Zhou, Z. Ma, Q. Wen, X. Wang, L. Sun, and R. Jin. FEDformer: Frequency enhanced decomposed transformer for long-term series forecasting. In ICML, volume 162, pages 27268– 27286, 2022

  28. [37]

    T. Zhou, P. Niu, X. Wang, L. Sun, and R. Jin. One fits all: Power general time series analysis by pretrained LM. In NeurIPS, 2023. A Datasets Description The detailed information for the dataset is presented below:

  29. [38]

    The Weather dataset [29] includes one-year records from 21 meteorological stations located in Germany, with a sampling rate of 10 minutes

  30. [39]

    It contains the hourly data recorded by the 762 sensors of San Francisco freeways from 2015 to 2016

    The traffic dataset [ 29] describes the road occupancy rates. It contains the hourly data recorded by the 762 sensors of San Francisco freeways from 2015 to 2016

  31. [40]

    The Electricity dataset [29] comprises two-year records of electricity consumption from 321 customers, measured at a 1-hour sampling rate

  32. [41]

    The influenza-like illness (ILI) dataset [29] contains records of patients experiencing severe influenza with complications

  33. [42]

    oil temperature

    The Electricity Transformer Temperature (ETT; An indicator reflective of long-term electric power deployment) [35] benchmark is comprised of two years of data, sourced from two counties in China. It comprises two hourly-level datasets (ETTh) and two 15-minute-level datasets (E...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.