Pith. sign in

REVIEW 5 major objections 6 minor 3 cited by

Unifying Prediction and Explanation in Time-Series Transformers via Shapley-based Pretraining

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read ShapTST trains a time-series transformer to output Shapley-value explanations in the same forward pass as its prediction, trading post-hoc repeated inference for a one-time pretraining cost.

desk verdict Plausible amortized Shapley for time-series transformers, but the training signal for the explainer head is missing from the paper. read the letter →

arxiv 2501.15070 v1 pith:HBQPO5R4 submitted 2025-01-25 cs.LG

Qisen Cheng , Jinming Xing , Chang Xue , Xiaoran Yang This is my paper · ORCID
classification cs.LG
keywords time-seriestransformerShapleyvaluesexplainabilityamortizedexplanationpretrainingcontrastivelearningmaskingrobustness
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes ShapTST, a training scheme that makes a time-series transformer output both its prediction and a Shapley-value explanation in one forward pass. Shapley values are per-input credit-assignment scores that say how much each time step and feature contributed to a prediction; they are normally computed after training by feeding many masked copies of the sample through the model. ShapTST instead trains an extra explainer head during a pretraining stage to imitate that repeated-inference process, so at deployment the explanation comes out with the prediction. On eight public datasets the approach matches or improves prediction accuracy, yields faithfulness scores comparable to TimeSHAP, and measures explanations 4 to 5 times faster than TimeSHAP. The practical target is safety-critical applications where every prediction must be accompanied by a reason without slowing down the system.

What carries the argument

The key machinery is the explainer head paired with a Shapley-estimation loss and three masking functions. The loss supervises the head to reproduce the marginal contribution $v_{x,y}(S)-v_{x,y}(\emptyset)$ of every masked subset, amortizing the repeated inference that post-hoc methods perform at test time; the masks $m_T$, $m_D$, $m_C$ define the subsets at time, feature, and cell granularity. A pretraining objective that sums InfoNCE, cross-entropy, and a KL-divergence between original and perturbed predictions ties this explanation learning to representation learning, and the efficiency-gap correction forces the estimated Shapley values to sum to the full prediction difference.

What would settle it

Compare ShapTST's per-cell Shapley estimates against exact Shapley values computed with a very high-sample post-hoc method on held-out masks that are rare in pretraining—for example masks covering large contiguous time blocks or unusual feature combinations. If faithfulness on those masks falls well below the Table III averages, the amortized estimator fits the pretraining mask distribution rather than learning the true Shapley function.

Watch

Extended reading notes

Core claim

The central claim is that Shapley explanations for a time-series transformer can be learned rather than computed post-hoc. ShapTST adds an explainer head $\phi_\gamma(x,y)\in\mathbb{R}^{D\times T}$ trained with the FastSHAP-style loss $\mathcal{L}_\gamma = \mathbb{E}_{p(x)}\mathbb{E}_{U(y)}\mathbb{E}_{p(S)}[(v_{x,y}(S)-v_{x,y}(\emptyset)-\mathbf{1}_S^\top \phi_\gamma(x,y))^2]$, where $S$ ranges over subsets generated by time-, feature-, and cell-level masking functions (Eqs. (4)-(6)). After training, the model produces predictions and per-cell Shapley values simultaneously, with the estimates refined by the efficiency-gap correction of Eq. (3). The paper reports that on eight datasets this yields competitive predictions, robustness to noise, faithfulness at least on par with TimeSHAP, and a 4-5x reduction in explanation evaluation time.

Load-bearing premise

The load-bearing premise is that the explainer head, trained on a sampled set of masked subsets, gives correct Shapley values for all masks used at test time—including masks that never appeared in pretraining—which requires the pretraining mask distribution and the uninformative replacement values to match how information is removed in deployment.

Editorial extensions

If this is right

  • Explanations no longer require repeated forward passes at deployment: after pretraining, ShapTST computes predictions and Shapley values simultaneously, with a measured 4-5x reduction in explanation evaluation time versus TimeSHAP on the tested datasets.
  • The amortized estimator makes explanation cost independent of the number of sampled subsets, so at high explanation frequencies the savings grow and may make per-request Shapley explanations practical in streaming or safety-critical settings.
  • The pretraining objective doubles as a representation learner: ShapTST improves over its vanilla TST baseline on several classification and regression benchmarks, with the largest AUROC gain (2.7%) on the larger Gas Sensor dataset, and improves robustness to Gaussian noise when a noisy feature's Shapley target is regularized to zero.
  • Because the framework is attached as an extra head plus losses, it can be layered onto existing transformer backbones; the paper demonstrates it with TST, PatchTST, and ViTST with results on par with the originals.

Reading between the lines

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

  • The paper leaves the uninformative baseline $\tilde{x}$ and the mask sampling distribution $p(S)$ unspecified; an editorially natural reading is that the amortized Shapley values are only guaranteed to be accurate for masks the pretraining distribution represents, so faithfulness on arbitrary held-out masks is an open empirical question.
  • If the estimator generalizes, the same 'pretrain an explainer head with a Shapley loss' recipe could be applied to other sequence models (speech, medical monitoring), where the savings compound because TimeSHAP-style repeated inference is per request.
  • The controllable Shapley target in Section II-D suggests a mechanism for injecting prior knowledge into a model—setting a known-spurious feature's Shapley contribution to zero—which could be tested as a general debiasing tool beyond noise robustness.
  • Since the pretraining uses only the input signal and labels, the approach could be extended to self-supervised setups with unlabeled time series, using the KL term between original and perturbed predictions as an explanation-aware augmentation.
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

5 major / 6 minor

Summary. The paper proposes ShapTST, a time-series transformer framework with an additional explainer head intended to output Shapley-value-based explanations for time steps, features, and individual cells. The stated design is to train this head with a FastSHAP-style regression loss while pretraining the backbone with a combination of InfoNCE, cross-entropy, and KL-divergence losses, and then to use the head at inference to provide explanations in a single forward pass. The authors report competitive or improved prediction performance on several classification and regression datasets, a 4-5x reduction in explanation evaluation time compared with TimeSHAP, faithfulness scores on par with TimeSHAP, and improved robustness to noise via a Shapley-based regularization term. The central proposal is that explanation and prediction can be unified during training, eliminating expensive post-hoc Shapley computation.

Significance. If the explainer head genuinely learns the Shapley values of the trained model, the amortized, single-pass explanation mechanism would be a valuable contribution for safety-critical time-series applications, and the idea of integrating Shapley-based supervision into pretraining is worth serious consideration. The paper usefully identifies an efficiency limitation of post-hoc Shapley computation and proposes a concrete alternative architecture. However, the manuscript currently leaves the core training mechanism for the explainer head unspecified, and several experimental details are missing, so the central claims about explanation quality and single-pass efficiency are not yet fully supported. The paper does not provide code, error bars, or hyperparameter values, which further limits verification. The potential significance is real, but the present version does not establish it.

major comments (5)
  1. [II-C, Eq. (7)] The pretraining loss L_P in Eq. (7) has no Shapley regression term and no dependence on the explainer parameters gamma. Eq. (1) defines L_gamma as the supervised regression of the head to the marginal contributions v_{x,y}(S)-v_{x,y}(empty)-S^T phi_gamma(x,y), but Eq. (7) consists only of InfoNCE, CE, and KL terms. The text states that the KL term 'allows for Shapley value estimation,' yet that term compares predictions of the original and perturbed inputs rather than the explainer's additive attribution to the value difference. The manuscript never specifies how phi_gamma is trained toward Shapley values. If the intended procedure is a two-stage schedule (pretrain with Eq. (7), then fine-tune with Eq. (2)), the schedule and relative weights must be stated; if Eq. (1) is optimized jointly with the backbone, the targets v_{x,y}(S) change as f_w changes, so the standard FastSHAP convergence argument does not apply directly. This is load-bearing because the faithfulness results in Table III can only be interpreted as measuring Shapley accuracy if the head actually received supervision toward Shapley values.
  2. [II-A, Eq. (3) and III-D] The claim that the model 'simultaneously generates predictions and Shapley value estimations in a single forward pass' is not consistent with Eq. (3), which refines the head outputs by adding v_{x,y}(1) - v_{x,y}(empty) - 1^T phi_gamma(x,y)/(D*T). Computing this refinement requires evaluating v_{x,y}(empty), the model output on a fully masked input, which is not already available from the prediction pass. The paper does not explain how v_{x,y}(empty) is obtained without an additional forward pass, or whether the refinement is omitted at inference despite being part of the method description. The speed comparison in Table II therefore needs a clear statement of what exactly is measured, including whether this extra evaluation is included for ShapTST and for TimeSHAP.
  3. [II-B, Eqs. (4)-(6)] The masking functions m_T, m_D, and m_C all rely on an 'uninformative values' matrix x~, but the paper never defines how x~ is chosen. For Shapley values, the replacement baseline is part of the definition of the value function v_{x,y}(S), so different baselines (zero, mean, per-feature median, or a learned value) lead to different explanations. Similarly, the subset distribution p(S) and the sampling weights over time, feature, and cell masks are not specified. Without these choices, the equivalence between ShapTST's head outputs and TimeSHAP's Shapley values cannot be assessed, and the faithfulness comparison in Table III is not reproducible.
  4. [III, Table I and Section III-A] The abstract states that the method is validated on eight public datasets, but Table I contains six distinct datasets (AReM and Gas sensor each appear under two tasks). The paper also reports no error bars, confidence intervals, or number of random seeds for any of the prediction, faithfulness, or robustness results, and the hyperparameters alpha_CE, alpha_KL, lambda, and p_mask are introduced but never given numerical values. These omissions make it impossible to determine whether the observed prediction improvements and the robustness effects are statistically meaningful, and they prevent replication of the experiments.
  5. [III-C, Table III] The faithfulness score is referenced to [36] but the exact computation used to produce Table III is not described in the manuscript. It is not stated which masking procedure is used for evaluation, how the values are aggregated over time steps and features, or how the score is normalized. Since faithfulness is the main quantitative evidence for explanation quality, the reader cannot verify whether the reported numbers (e.g., 0.650 for TimeSHAP vs. 0.688 for ShapTST on AReM) are computed on an apples-to-apples basis. Please specify the faithfulness definition and the evaluation protocol.
minor comments (6)
  1. [I and III-C] The name 'TimeShap' appears in the introduction and in the caption of Table III, while the reference is TimeSHAP [19]; please use a consistent spelling.
  2. [III] The dataset name 'EGG eye state' should be 'EEG eye state'.
  3. [III-D, Table II] Table II reports times without units and does not state how many samples or test instances were used for the timing comparison, nor whether the hardware and batch settings were identical for both methods.
  4. [II-B, Eq. (6)] In Eq. (6), S_T and S_D are vectors of different lengths, and the notation S_T * S_D is ambiguous; please clarify whether this is an outer product or a broadcasting rule.
  5. [II-D] The Shapley-based regularization uses a target y_target (e.g., zero) but the paper does not explain how a user would choose this target for a given feature or time step, or how the L2 loss is scaled relative to the other losses.
  6. [II-A, Eq. (2)] The total loss L_DT in Eq. (2) sums L_gamma over time, feature, and cell levels, but it is not explained how the three losses are weighted relative to each other and to L_w, nor how these weights are chosen in the experiments.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor construct-level circularity: the faithfulness evaluation recapitulates the explainer's own training objective; prediction and speed claims remain independent.

  1. self definitional [Section II-A Eq. (1) and Section III-C Table III]
    "Lγ = Ep(x)EU (y)Ep(S) (vx,y(S) − vx,y(∅) − S^T ϕγ(x, y))^2 (1) ... To evaluate the quality of generated explanations, we compare the faithfulness [36] of our generated explanations with those from TimeSHAP [19]. Table III reports the faithfulness scores averaged over the time steps and features for all 4 classification datasets, showing the quality of our explanations are on par with or better than TimeSHAP."

    The explainer head is supervised by Eq. (1) to minimize, over sampled masks S, the squared error between S^T ϕγ(x,y) and the model's own value differences v(S) − v(∅). A removal-based faithfulness metric rewards essentially the same agreement between explanations and model output changes under masking. Thus Table III's 'on par with or better than TimeSHAP' is largely a statement that the head fit its own regression target on the training mask distribution, not independent evidence of Shapley correctness for unseen masks. The prediction, speed, and robustness claims do not depend on this equivalence, so the circularity is partial and not central.

full rationale

No load-bearing self-citation chain is present: the method cites FastSHAP, SCARF, and TimeSHAP as external prior work, and no author-owned uniqueness theorem is invoked. The central efficiency claim (single-forward-pass explanation generation) is independent of whether the Shapley values are exactly correct, and the prediction and robustness results are benchmarked against external baselines. The one construct-level circularity is in the explanation-quality evaluation: because Eq. (1) trains the explainer to match the same model-value differences that faithfulness scores reward, the faithfulness comparison is partly a fit check. A separate, non-circular completeness issue is that Eq. (7), the stated pretraining loss, contains no Lγ term and no dependence on ϕγ, so the paper does not fully specify how the Shapley head is trained; this is a missing-derivation risk rather than circularity. Overall circularity burden is low.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard Shapley axioms, on FastSHAP's amortized estimation guarantee, on a masking baseline that must represent missing information, and on the proposed pretraining objective improving both prediction and explanation. Several hyperparameters are free and unreported.

free parameters (4)
  • alpha_CE
    Weight of the cross-entropy loss in pretraining Eq. (7); no value or tuning procedure is reported.
  • alpha_KL
    Weight of the KL-divergence loss in pretraining Eq. (7); no value or tuning procedure is reported.
  • lambda (time-step aggregation factor)
    Aggregation factor in the time-level masking Eq. (4); no value is reported and it changes explanation granularity.
  • p_mask (mask probability)
    Mask probability used for the noise-robustness regularization in Section II-D; the paper says it is determined by confidence measures but does not give the formula.
assumptions (4)
  • domain assumption Amortized Shapley estimation via Eq. (1) converges to the true Shapley values of the trained model.
    Relies on the FastSHAP result that regression against sampled marginal contributions yields Shapley values; invoked in Section II-A.
  • domain assumption Masking with uninformative values x~ represents valid removal of information from the input.
    Section II-B, Eqs. (4)-(6); Shapley values depend on the baseline, but x~ is never specified and no check of on-manifold behavior is provided.
  • ad hoc to paper The InfoNCE, CE, and KL pretraining objective Eq. (7) improves both prediction and explanation quality.
    Section II-C; this is the paper's proposed mechanism, presented without an ablative decomposition or theoretical justification.
  • standard math Shapley value axioms, including efficiency, apply to the model's characteristic function v(x, y, S).
    Used implicitly in the efficiency-gap correction Eq. (3).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unifying Prediction and Explanation in Time-Series Transformers via Shapley-based Pretraining." pith.science (2026). https://pith.science/paper/HBQPO5R4

@misc{pith2026250115070,
  author       = {Pith},
  title        = {Pith review of: Unifying Prediction and Explanation in Time-Series Transformers via Shapley-based Pretraining},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HBQPO5R4}},
  note         = {Machine review of arXiv:2501.15070}
}
read the original abstract

In this paper, we propose ShapTST, a framework that enables time-series transformers to efficiently generate Shapley-value-based explanations alongside predictions in a single forward pass. Shapley values are widely used to evaluate the contribution of different time-steps and features in a test sample, and are commonly generated through repeatedly inferring on each sample with different parts of information removed. Therefore, it requires expensive inference-time computations that occur at every request for model explanations. In contrast, our framework unifies the explanation and prediction in training through a novel Shapley-based pre-training design, which eliminates the undesirable test-time computation and replaces it with a single-time pre-training. Moreover, this specialized pre-training benefits the prediction performance by making the transformer model more effectively weigh different features and time-steps in the time-series, particularly improving the robustness against data noise that is common to raw time-series data. We experimentally validated our approach on eight public datasets, where our time-series model achieved competitive results in both classification and regression tasks, while providing Shapley-based explanations similar to those obtained with post-hoc computation. Our work offers an efficient and explainable solution for time-series analysis tasks in the safety-critical applications.

Figures

Figures reproduced from arXiv: 2501.15070 by the authors.

Figure 1
Figure 1. The ShapTST framework. It involves 1) a Shaley-based pre-training stage, where TST is trained on both original and perturbed data to learn robust [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Maskings for the training of ShapTST 2) Feature level masking: The feature level masking func￾tion is used to replace a random subset of features with uninformative values in a given matrix x. Here, the subset of features is denoted as SD ∈ {0, 1} D, where D represents the number of features. The matrix xe has the same shape as x and contains uninformative values. Since features are often independent, there is no ne… view at source ↗
Figure 5
Figure 5. Time and feature level explanation examples (AReM). Adjacent 4 [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figures from the paper (2 more)
Figure 3
Figure 3. Figure 3: Performance on AReM dataset with added noise. The blue line is [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 4
Figure 4. Figure 4: Time level explanation example of AReM Dataset. Adjacent 4 time [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Revisiting LRP: Positional Attribution as the Missing Ingredient for Transformer Explainability

    cs.LG 2025-06 conditional novelty 5.0 of 10

    PA-LRP extends Layer-wise Relevance Propagation to attribute relevance to positional encodings in Transformers, improving faithfulness of explanations.

  2. FGATT: A Robust Framework for Wireless Data Imputation Using Fuzzy Graph Attention Networks and Transformer Encoders

    cs.LG 2024-12 reject novelty 4.0 of 10

    FGATT, a fuzzy graph attention network paired with a Transformer encoder, is claimed to improve missing-data imputation on two SWaT wireless sensor datasets.

  3. Comparative Analysis of Pooling Mechanisms in LLMs: A Sentiment Analysis Perspective

    cs.CL 2024-11 conditional novelty 2.0 of 10

    An empirical comparison finds that Mean pooling works best for BERT sentiment classification while Weighted Sum works best for GPT-2, though differences are small and reported without error bars.

Reference graph

Works this paper leans on

36 extracted references · 24 canonical work pages · cited by 3 Pith papers

  1. [36]

    Evaluating and aggregating feature-based model explanations,

    U. Bhatt, A. Weller, and J. M. Moura, “Evaluating and aggregating feature-based model explanations,” arXiv preprint arXiv:2005.00631 , 2020

    arXiv 2005

  2. [1]

    Financial time series forecasting with deep learning: A systematic literature review: 2005–2019,

    O. B. Sezer, M. U. Gudelek, and A. M. Ozbayoglu, “Financial time series forecasting with deep learning: A systematic literature review: 2005–2019,” Applied soft computing , vol. 90, p. 106181, 2020

    work page 2005

  3. [2]

    Augmented time regularized generative adversarial network (atr-gan) for data aug- mentation in online process anomaly detection,

    Y . Li, Z. Shi, C. Liu, W. Tian, Z. Kong, and C. B. Williams, “Augmented time regularized generative adversarial network (atr-gan) for data aug- mentation in online process anomaly detection,” IEEE Transactions on Automation Science and Engineering , vol. 19, no. 4, pp. 3338–3355, 2021

    work page 2021

  4. [3]

    Deepsense: A unified deep learning framework for time-series mobile sensing data processing,

    S. Yao, S. Hu, Y . Zhao, A. Zhang, and T. Abdelzaher, “Deepsense: A unified deep learning framework for time-series mobile sensing data processing,” in Proceedings of the 26th international conference on world wide web , pp. 351–360, 2017

    work page 2017

  5. [4]

    Analysis of lifting posture by two inertial measurement units and a classification model based on a convolutional neural network,

    J. Liu, X. Yang, J.-C. Hong, and H. Iwata, “Analysis of lifting posture by two inertial measurement units and a classification model based on a convolutional neural network,” in 2024 10th IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics (BioRob), pp. 383–388, IEEE, 2024

    work page 2024

  6. [5]

    Attention-based cnn capturing eeg recording’s average voltage and local change,

    L. Yi and X. Qu, “Attention-based cnn capturing eeg recording’s average voltage and local change,” in International Conference on Human- Computer Interaction, pp. 448–459, Springer, 2022

    work page 2022

  7. [6]

    Estimation of energy and time usage in 3d printing with multimodal neural network,

    Q. Cheng, C. Zhang, and X. Shen, “Estimation of energy and time usage in 3d printing with multimodal neural network,” in 2022 4th International Conference on Frontiers Technology of Information and Computer (ICFTIC), pp. 900–903, IEEE, 2022

    work page 2022

  8. [7]

    Predicting melt-crystal interface position and shape during the manufacturing process of single crystal via explainable machine learning models,

    C. Zhang and Q. Cheng, “Predicting melt-crystal interface position and shape during the manufacturing process of single crystal via explainable machine learning models,” in IOP Conference Series: Materials Science and Engineering, vol. 1258, p. 012029, IOP Publishing, 2022

    work page 2022

Show all 36 references
  1. [8]

    6-4: Deep learning for classification of repairable defects in display panels using multi-modal data,

    K. Balakrishnan, Q. Cheng, J. Lee, D. Jeong, E. Kim, and J. Kim, “6-4: Deep learning for classification of repairable defects in display panels using multi-modal data,” in SID Symposium Digest of Technical Papers, vol. 54, pp. 58–61, Wiley Online Library, 2023

    2023

  2. [9]

    Deep learning for time series classification: a review,

    H. Ismail Fawaz, G. Forestier, J. Weber, L. Idoumghar, and P.-A. Muller, “Deep learning for time series classification: a review,” Data mining and knowledge discovery, vol. 33, no. 4, pp. 917–963, 2019

    2019

  3. [10]

    Trans- formers in time series: A survey,

    Q. Wen, T. Zhou, C. Zhang, W. Chen, Z. Ma, J. Yan, and L. Sun, “Trans- formers in time series: A survey,” arXiv preprint arXiv:2202.07125 , 2022

    2022 arXiv

  4. [11]

    Enhancing eye-tracking performance through multi-task learning transformer,

    W. Li, N. Zhou, and X. Qu, “Enhancing eye-tracking performance through multi-task learning transformer,” in International Conference on Human-Computer Interaction , pp. 31–46, Springer, 2024

    2024

  5. [12]

    Benchmark- ing deep learning interpretability in time series predictions,

    A. A. Ismail, M. Gunady, H. Corrada Bravo, and S. Feizi, “Benchmark- ing deep learning interpretability in time series predictions,” Advances in neural information processing systems, vol. 33, pp. 6441–6452, 2020

    2020

  6. [13]

    32-1: Improving qd backplane defect image generation using automatic masking in diffusion models,

    Z. Pan, R. Shenoy, K. Balakrishnan, Q. Cheng, J. Lee, Y . Jeon, D. Jeong, and J. Kim, “32-1: Improving qd backplane defect image generation using automatic masking in diffusion models,” in SID Symposium Digest of Technical Papers, vol. 55, pp. 409–412, Wiley Online Library, 2024

    2024

  7. [14]

    Short-term traffic condition pre- diction based on multi-source data fusion,

    X. Deng, H. Zhou, X. Yang, and C. Ye, “Short-term traffic condition pre- diction based on multi-source data fusion,” in International Conference on Data Mining and Big Data , pp. 327–335, Springer, 2021

    2021

  8. [15]

    Local interpretable model- agnostic explanations for music content analysis.,

    S. Mishra, B. L. Sturm, and S. Dixon, “Local interpretable model- agnostic explanations for music content analysis.,” in ISMIR, vol. 53, pp. 537–543, 2017

    2017

  9. [16]

    A unified approach to interpreting model predictions,

    S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” Advances in neural information processing systems, vol. 30, 2017

    2017

  10. [17]

    Improving kernelshap: Practical shapley value estimation using linear regression,

    I. Covert and S.-I. Lee, “Improving kernelshap: Practical shapley value estimation using linear regression,” in International Conference on Artificial Intelligence and Statistics , pp. 3457–3465, PMLR, 2021

    2021

  11. [18]

    J. D. Hamilton, Time series analysis . Princeton university press, 2020

    2020

  12. [19]

    Time- shap: Explaining recurrent models through sequence perturbations,

    J. Bento, P. Saleiro, A. F. Cruz, M. A. Figueiredo, and P. Bizarro, “Time- shap: Explaining recurrent models through sequence perturbations,” in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pp. 2565–2573, 2021

    2021

  13. [20]

    A survey on lstm memristive neural network architectures and applications,

    K. Smagulova and A. P. James, “A survey on lstm memristive neural network architectures and applications,” The European Physical Journal Special Topics, vol. 228, no. 10, pp. 2313–2324, 2019

    2019

  14. [21]

    Fastshap: Real-time shapley value estimation,

    N. Jethani, M. Sudarshan, I. C. Covert, S.-I. Lee, and R. Ranganath, “Fastshap: Real-time shapley value estimation,” in International Con- ference on Learning Representations , 2021

    2021

  15. [22]

    What makes for good views for contrastive learning?,

    Y . Tian, C. Sun, B. Poole, D. Krishnan, C. Schmid, and P. Isola, “What makes for good views for contrastive learning?,” Advances in neural information processing systems , vol. 33, pp. 6827–6839, 2020

    2020

  16. [23]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

    2017

  17. [24]

    Specaugment: A simple data augmentation method for automatic speech recognition,

    D. S. Park, W. Chan, Y . Zhang, C.-C. Chiu, B. Zoph, E. D. Cubuk, and Q. V . Le, “Specaugment: A simple data augmentation method for automatic speech recognition,” arXiv preprint arXiv:1904.08779 , 2019

    1904 arXiv

  18. [25]

    Scarf: Self-supervised contrastive learning using random feature corruption,

    D. Bahri, H. Jiang, Y . Tay, and D. Metzler, “Scarf: Self-supervised contrastive learning using random feature corruption,” arXiv preprint arXiv:2106.15147, 2021

    2021 arXiv

  19. [26]

    Adversarial attacks on time series,

    F. Karim, S. Majumdar, and H. Darabi, “Adversarial attacks on time series,” IEEE transactions on pattern analysis and machine intelligence , vol. 43, no. 10, pp. 3309–3320, 2020

    2020

  20. [27]

    72-3: Deep learning based visual defect detection in noisy and imbalanced data,

    Q. Cheng, S. Qu, and J. Lee, “72-3: Deep learning based visual defect detection in noisy and imbalanced data,” in SID Symposium Digest of Technical Papers, vol. 53, pp. 971–974, Wiley Online Library, 2022

    2022

  21. [28]

    Human activity recognition using multisensor data fusion based on reservoir computing,

    F. Palumbo, C. Gallicchio, R. Pucci, and A. Micheli, “Human activity recognition using multisensor data fusion based on reservoir computing,” Journal of Ambient Intelligence and Smart Environments , vol. 8, no. 2, pp. 87–107, 2016

    2016

  22. [29]

    Eeg eye state identifica- tion using incremental attribute learning with time-series classification,

    T. Wang, S.-U. Guan, K. L. Man, and T. Ting, “Eeg eye state identifica- tion using incremental attribute learning with time-series classification,” Mathematical Problems in Engineering , vol. 2014, 2014

    2014

  23. [30]

    Online decorrelation of humidity and temperature in chemical sensors for continuous monitoring,

    R. Huerta, T. Mosqueiro, J. Fonollosa, N. F. Rulkov, and I. Rodriguez- Lujan, “Online decorrelation of humidity and temperature in chemical sensors for continuous monitoring,” Chemometrics and Intelligent Lab- oratory Systems, vol. 157, pp. 169–176, 2016

    2016

  24. [31]

    Condition monitoring of a complex hydraulic system using multivariate statistics,

    N. Helwig, E. Pignanelli, and A. Sch ¨utze, “Condition monitoring of a complex hydraulic system using multivariate statistics,” in 2015 IEEE International Instrumentation and Measurement Technology Conference (I2MTC) Proceedings, pp. 210–215, IEEE, 2015

    2015

  25. [32]

    Monash uni- versity, uea, ucr time series extrinsic regression archive,

    C. W. Tan, C. Bergmeir, F. Petitjean, and G. I. Webb, “Monash uni- versity, uea, ucr time series extrinsic regression archive,” arXiv preprint arXiv:2006.10996, 2020

    2006 arXiv

  26. [33]

    A transformer-based framework for multivariate time series representation learning,

    G. Zerveas, S. Jayaraman, D. Patel, A. Bhamidipaty, and C. Eickhoff, “A transformer-based framework for multivariate time series representation learning,” in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pp. 2114–2124, 2021

    2021

  27. [34]

    A time series is worth 64 words: Long-term forecasting with transformers,

    Y . Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam, “A time series is worth 64 words: Long-term forecasting with transformers,” arXiv preprint arXiv:2211.14730, 2022

    2022 arXiv

  28. [35]

    Time series as images: Vision transformer for irregularly sampled time series,

    Z. Li, S. Li, and X. Yan, “Time series as images: Vision transformer for irregularly sampled time series,” Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

Pith tools

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