Pith. sign in

REVIEW 5 major objections 5 minor 3 cited by

Camel: Energy-Aware LLM Inference on Resource-Constrained Devices

T0 review · 5 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The full-text paper claims PatchECG, a masked-training patch classifier, detects arrhythmias from digitized multi-layout ECG images without interpolation, reaching 0.893 on real 12-lead images; the abstract's Camel energy-management claim a

desk verdict The submission is two unrelated papers stapled together: the abstract promises an LLM energy-management result the body never mentions, while the body is a competent ECG-arrhythmia paper that deserves a clean resubmission, not review under this wrapper. read the letter →

arxiv 2508.09173 v1 pith:IELK4D7J submitted 2025-08-07 cs.NI

Hao Xu , Long Peng , Shezheng Song , Xiaodong Liu , Ma Jun , Shasha Li , Jie Yu , Xiaoguang Mao This is my paper · ORCID
classification cs.NI
keywords Camelenergy-delayproductECGimagedigitizationarrhythmiadetectionmaskedtrainingpartialblackoutatrialfibrillation
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

This manuscript has two incompatible layers. The abstract and title claim Camel, an energy-management framework that reduces energy-delay product for edge LLM inference by 12.4%-29.9%, but the full text contains no Camel design, experiment, or result. The body is a complete paper about PatchECG, an interpolation-free arrhythmia classifier for digitized multi-layout ECG images, trained on PTB-XL and validated on simulated layouts and 400 real hospital images; that is what a sympathetic reader can take as the paper's real contribution. If PatchECG's numbers hold, it would make decades of paper ECG archives usable for automated diagnosis in settings where only images exist. The abstract's Camel claim must be treated separately, since nothing in the provided text substantiates it.

What carries the argument

The machinery is the adaptive variable-block masking plus patch-drop attention: each lead receives at most one contiguous missing block of uniformly random start and length, the signal is cut into fixed-length patches, fully missing patches are discarded rather than imputed, and a learned Segment-Shuffle-Stitch reordering plus lead/time embeddings lets a transformer attend across leads and time regardless of layout. This lets the model process arbitrary lead counts and lengths without interpolation.

What would settle it

Re-train PatchECG using masks sampled from the actual layout schedules—3×4 leads starting at 0, 2.5, 5.0, and 7.5 s; 6×2 leads at 0 and 5 s—allowing multiple missing blocks per lead, and evaluate on the same external cohort. If the 0.893 (12×1) and 0.778 (mixed-layout) values drop materially, the single-block uniform mask in Eq. (1) did not represent real digitization structure.

Watch

Extended reading notes

Core claim

PatchECG's central claim is that arrhythmia detection can be made robust to the messy output of ECG image digitization—where leads start at different times (2.5 s or 5 s offsets) and contiguous chunks of signal are lost ('partial blackout')—without filling in or interpolating missing data. The model divides each lead into fixed-length patches, drops patches that are entirely missing, marks the rest with a binary observed/missing indicator, and passes them through a patch encoder followed by a learned, layout-agnostic attention module. Trained on PTB-XL and tested under seven simulated layouts, it reports an average AUROC of about 0.835; on 400 real hospital ECG images it reports 0.778 for at

Load-bearing premise

The results stand on the assumption that the single random missing-segment-per-lead used in training matches the timing shifts and blackouts that occur when real multi-layout ECG images are digitized, and that the manually digitized hospital signals preserve enough diagnostic information for the reported accuracy scores to reflect model ability rather than digitization noise.

Editorial extensions

If this is right

  • If PatchECG's numbers hold, digitized ECG images from any layout can be classified directly as signals, so interpolation noise is never introduced.
  • Existing signal encoders—including a large pre-trained ECG foundation model—can be dropped into PatchECG as patch encoders, letting them work on arbitrary lead counts and lengths.
  • On real hospital data, the model's AUROC is higher on complete 12×1 layouts (0.893) than on mixed layouts (0.778), suggesting layout completeness is a performance driver.
  • Attention scores align with cardiologist-selected patches at a rate approaching inter-clinician agreement (up to 36.0% top-20 overlap vs 41.8% between doctors), supporting signal-grounded interpretability.
  • At 21.72M parameters and about 1.09 ms per sample, the model is light enough for real-time use alongside the proposed digitization workflow.

Reading between the lines

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

  • A reader should check the abstract's 'approximately 0.835 average' against Table 3: the seven listed layouts for the Net1D encoder average about 0.831, and 0.835 appears as the Random-layout value.
  • A fair attribution test would fine-tune ECGFounder under the same masking and calibration before comparing; the reported margins may mix the contributions of masking, architecture, and pre-training.
  • The training mask allows one contiguous missing block per lead, while digitized 3×4 and 6×2 images have lead groups starting at fixed 2.5s/5s offsets; testing with masks that follow those schedules would stress the robustness claim.
  • The manual digitization fidelity metrics are reported as placeholders, so the external-hospital accuracy numbers should be read as conditional on digitization quality until paired electronic-ground-truth images are evaluated.
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 submission arXiv:2508.09173 presents an abstract claiming an LLM inference energy-management framework, 'Camel', which optimizes GPU frequency and batch size on the NVIDIA Jetson AGX Orin and reduces energy-delay product (EDP) by 12.4%–29.9% relative to the default configuration. The full text, however, is a different paper: 'Masked Training for Robust Arrhythmia Detection from Digitalized Multiple Layout ECG Images', proposing PatchECG. The body contains no Camel method section, no definition of EDP, no GPU frequency/batch-size search, no latency or energy measurements, and no comparison against a default configuration. Instead, it reports a patch-based ECG classifier trained on PTB-XL, evaluated on simulated layouts and on 400 external Chaoyang Hospital ECG images, with an overall AUROC of 0.778 and 0.893 on a 12×1 subset. The stated central claim of the submitted paper is therefore unsupported by the manuscript's content, while the body's PatchECG results constitute a separate contribution that cannot be evaluated as evidence for the abstract claim.

Significance. If the abstract's claim were supported, a 12.4%–29.9% EDP reduction on a Jetson-class device would be a practically valuable result for edge LLM deployment. But as submitted, no experiment, table, or figure pertains to EDP, GPU frequency, batch size, energy, or latency; the central claim is entirely unsubstantiated. The PatchECG content, considered independently, addresses a relevant clinical problem—directly modeling asynchronous and partially missing multi-lead ECG signals without interpolation—and has strengths: public data and code, comparison against several baselines, an external real-hospital cohort, and a quantitative interpretability evaluation against cardiologists. However, this is a different model, task, and hardware platform from the abstract's Camel contribution, and several key tables contain placeholder values. The manuscript in its current form cannot be accepted or meaningfully revised as a paper about energy-aware LLM inference.

major comments (5)
  1. [Abstract vs. full text] The abstract's central claim—Camel reduces EDP by 12.4%–29.9% by optimizing GPU frequency and batch size on Jetson AGX Orin—has no supporting content in the full text. The body is a complete ECG manuscript (PatchECG) with no mention of Camel, EDP, Jetson, GPU frequency, batch-size search, latency, energy, or default configuration. The footer even carries a different arXiv identifier (2508.09165v3). This is not a local gap; the paper's stated central result cannot be checked at all.
  2. [Tables 4 and 5] The external-validation comparison tables render all baseline cells as placeholder symbols (e.g., 'ECGFounder ����� � ����'), and Table 5 leaves SimMTM-KNN and SimMTM-SAITS as '–'. The claims that PatchECG surpasses ECGFounder by 0.111 (overall) and 0.190 (12×1) therefore cannot be verified from the submitted text. These numbers are load-bearing for the body's headline result, so the missing values prevent reproducibility of the main performance comparison.
  3. [Eq. (1) and Appendix A] The training-time masking model in Eq. (1) samples one contiguous block with uniform start and length independently per lead. Appendix A, by contrast, describes layout-induced asynchrony as fixed offsets: 3×4 leads start at 0/2.5/5/7.5 s and 6×2 leads start at 0/5 s, with whole leads shifted. The relationship between the random-mask training distribution and the actual layout shifts is not established. Without evidence that the simulated missing patterns match the test-time digitization process, the claim of 'consistent' AUROC across layouts is a correctness risk.
  4. [Eqs. (14)–(15) and following paragraph] Digitization quality for the Chaoyang cohort is reported as 'Avg SNR of � ���� dB and Avg PSNR of ����� dB', with the numeric values rendered as placeholders. The manuscript itself later describes the digitization quality as poor (case study, cases b and d), yet the external-cohort AUROC differences are the strongest evidence for the method. A load-bearing premise—that the digitized signals preserve enough diagnostic information—is therefore left unquantified.
  5. [PatchECG Achieve Better Results on Real Hospital ECG Images] The 0.893 AUROC is reported on a '12×1 subset' selected from the Chaoyang Hospital cohort, but no sample size, inclusion criteria, or multiple-comparison adjustment is given for this subset; the whole-cohort result is 0.778. Highlighting a post-hoc subset as a headline result without these details is misleading and precludes assessing whether the gain is stable or a selection artifact.
minor comments (5)
  1. [Code availability] Typographical errors: 'imolementation', 'repositorv', and 'from left to tight' in the Figure 7 caption should be corrected.
  2. [Eqs. (1)–(2)] The notation in Eq. (1) is corrupted in the rendered text (e.g., '��� ∼ U��� ��� � � ∼ U����� � − ����'), making the precise sampling distribution hard to parse.
  3. [Data availability / Code availability] Several URLs appear as placeholder strings ('��������') in both Data availability and Code availability; these need to be resolved.
  4. [Table 6] The interpretability comparison table also contains placeholder values ('���� ± ����'), and the model-vs-clinician agreement is based on only 20 samples; the reported percentages should be read with appropriate uncertainty.
  5. [Abstract] The abstract says 'seven simulated layout conditions'; Table 3 has seven columns, but the text names only six layouts plus 'Random'. This should be clarified.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: PatchECG's accuracy claims are measured on held-out and external test data rather than derived from the training setup by construction; self-citations are non-load-bearing. The abstract's Camel EDP claim is unsupported (missing support) but not circular.

full rationale

The PatchECG derivation chain is not circular. The model is trained on raw PTB-XL signals (Methods/Training Data) with a random contiguous-masking augmentation (Eq. 1) and is evaluated on (i) held-out PTB-XL test folds and (ii) an external Chaoyang Hospital AF cohort digitized via PaperECG (Results). The reported AUROC values (Tables 3-5) are empirical measurements on test samples, not algebraic consequences of the training loss or of the masking distribution. The masking strategy is a data-augmentation prior, not a fitted parameter later reported as a 'prediction', and the simulated layouts impose missing patterns different from the training samples. Self-citations exist (ECGFounder ref 13, Chaoyang refs 34-35, Holmes ref 31), but they are used as a baseline, a data source, and a backbone module, respectively; in each case the paper runs the cited artifact or uses the cohort rather than assuming a result from the citation, so none of these citations is load-bearing for the paper's conclusions. No uniqueness theorem is imported from the authors' prior work. Flagged omissions that are not circularity: the abstract's central claim that the 'Camel' framework reduces EDP by 12.4%-29.9% is not supported by any experiment in the full text, which instead describes an unrelated ECG study (PatchECG); this is missing support, not a circular reduction. Similarly, the digitization-fidelity SNR/PSNR values in Eqs. (14)-(15) are placeholders ('�� ���� dB', '����� dB'), an omitted result the authors themselves acknowledge is only an indirect check. These issues lower confidence in the submission but do not make the PatchECG results circular.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The body introduces no new physical or conceptual entities; PatchECG is a composition of known modules. The main cost is instead in hand-set simulation choices and domain assumptions listed above. No invented entity needs independent falsification.

free parameters (5)
  • Uniform mask start/length distribution for training masking (Eq. 1)
    The mask start position and length are drawn from uniform distributions whose support is not tied to the actual 2.5s/5s lead-onset offsets found in digitized layouts; the choice of distribution is a hand-set simulation of missingness.
  • Masked value sentinel
    Masked signal values are set to a constant to distinguish them from observed zero-valued signal points. The constant is chosen by hand.
  • S3 initial segment count, multiplier, layers = 4 segments, multiplier 2, 3 layers
    Following the original S3 paper's ablation (ref 32), not derived for this task.
  • Model hyperparameters (window, filters, transformer dim, depth, heads) = w=64, dim=768, depth=3, heads=8
    Set to baseline-like values or found by searching hyperparameters, as the text says 'we attempt to maximize performance by searching for the best hyperparameters ourselves'.
  • Focal loss alpha/gamma
    The paper introduces Focal Loss for class imbalance but never reports the alpha or gamma values.
assumptions (4)
  • domain assumption PTB-XL diagnostic subclass labels are accurate ground truth for arrhythmia classification
    The model is trained and tested on these labels without independent adjudication.
  • domain assumption Digitized signals from ecg-image-kit/CINC 2024 winner and PaperECG preserve clinically relevant morphology
    The external comparison relies on this. The paper's own SNR/PSNR values are low and the reference signals are themselves digitized, not ground truth (Section on digitization quality).
  • domain assumption The S3 learned shuffle and transformer attention scores capture diagnostically meaningful structure
    The interpretability claim is validated on only 20 samples against two cardiologists, with model agreement below inter-clinician agreement.
  • ad hoc to paper Layout-induced asynchrony is representable by one contiguous random mask per lead during training
    Eq. 1 introduces this masking model; the paper does not show that real 2.5s/5s lead shifts produce the same missing pattern as one contiguous random block per lead.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Camel: Energy-Aware LLM Inference on Resource-Constrained Devices." pith.science (2026). https://pith.science/paper/IELK4D7J

@misc{pith2026250809173,
  author       = {Pith},
  title        = {Pith review of: Camel: Energy-Aware LLM Inference on Resource-Constrained Devices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IELK4D7J}},
  note         = {Machine review of arXiv:2508.09173}
}
read the original abstract

Most Large Language Models (LLMs) are currently deployed in the cloud, with users relying on internet connectivity for access. However, this paradigm faces challenges such as network latency, privacy concerns, and bandwidth limits. Thus, deploying LLMs on edge devices has become an important research focus. In edge inference, request latency is critical as high latency can impair real-time tasks. At the same time, edge devices usually have limited battery capacity, making energy consumption another major concern. Balancing energy consumption and inference latency is essential. To address this, we propose an LLM inference energy management framework that optimizes GPU frequency and batch size to balance latency and energy consumption. By effectively managing the exploration-exploitation dilemma in configuration search, the framework finds the optimal settings. The framework was implemented on the NVIDIA Jetson AGX Orin platform, and a series of experimental validations were conducted. Results demonstrate that, compared to the default configuration, our framework reduces energy delay product (EDP) by 12.4%-29.9%, achieving a better balance between energy consumption and latency.

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. Seeing is Free, Speaking is Not: Uncovering the True Energy Bottleneck in Edge VLM Inference

    cs.CV 2026-07 conditional novelty 6.5 of 10

    On edge hardware, VLM decode uses 86–97% of energy and each output token costs 11–39× an input token, so output length—not visual tokens—dominates energy.

  2. LUVE : Latent-Cascaded Ultra-High-Resolution Video Generation with Dual Frequency Experts

    cs.CV 2026-02 conditional novelty 6.0 of 10

    A latent-cascaded video generation framework with dual frequency-split experts reports state-of-the-art 2K/4K video generation on VBench, FIDpatch, and human preference.

  3. HealthSLM-Bench: Benchmarking Small Language Models for Mobile and Wearable Healthcare Monitoring

    cs.AI 2025-09 conditional novelty 5.0 of 10

    Small language models rival large ones on several wearable health prediction tasks, with large efficiency gains, but suffer from class imbalance and poor calorie regression.

Reference graph

Works this paper leans on

38 extracted references · 33 canonical work pages · cited by 3 Pith papers

  1. [1]

    Mensah, G.A., Roth, G.A., and Fuster, V. (2019). The global burden of cardiovascular 678 diseases and risk factors: 2020 and beyond. American College of Cardiology Foundation 679 Washington, DC. 680 22

    work page 2019

  2. [2]

    Goldberger, A.L., Amaral, L.A., Glass, L., Hausdorff, J.M., Ivanov, P .C., Mark, R.G., Mietus,681 J.E., Moody, G.B., Peng, C.K., and Stanley, H.E. (2000). PhysioBank, PhysioToolkit, and Phy-682 sioNet: Components of a new research resource for complex physiologic signals. Circulation 683 ��� , e215–e220. 684

    work page 2000

  3. [3]

    Sangha, V., Mortazavi, B.J., Haimovich, A.D., Ribeiro, A.H., Brandt, C.A., Jacoby, D.L., 685 Schulz, W.L., Krumholz, H.M., Ribeiro, A.L.P ., and Khera, R. (2022). Automated multilabel 686 diagnosis on electrocardiographic images and signals. Nature communications ��, 1583. 687

    work page 2022

  4. [4]

    (2025).688 Clinically meaningful interpretability of an ai model for ecg classification

    Gliner, V., Levy, I., Tsutsui, K., Acha, M.R., Schliamser, J., Schuster, A., and Y aniv, Y . (2025).688 Clinically meaningful interpretability of an ai model for ecg classification. NPJ Digital Medicine 689 �, 109. 690

    work page 2025

  5. [5]

    Y ang, K., Hong, M., Zhang, J., Luo, Y ., Zhao, S., Zhang, O., Yu, X., Zhou, J., Y ang, L., Zhang,691 P . et al. (2025). Ecg-lm: understanding electrocardiogram with a large language model. 692 Health Data Science �, 0221. 693

    work page 2025

  6. [6]

    Krones, F ., Walker, B., Lyons, T., and Mahdi, A. (2024). Combining hough transform 694 and deep learning approaches to reconstruct ecg signals from printouts. arXiv preprint 695 arXiv:2410.14185. 696

    arXiv 2024

  7. [7]

    Fortune, J.D., Coppa, N.E., Haq, K.T., Patel, H., and Tereshchenko, L.G. (2022). Digitizing 697 ecg image: A new method and open-source software code. Computer methods and programs 698 in biomedicine ��� , 106890. 699

    work page 2022

  8. [8]

    Sau, A., Zeidaabadi, B., Patlatzoglou, K., Pastika, L., Ribeiro, A.H., Sabino, E., Peters, 700 N.S., Ribeiro, A.L.P ., Kramer, D.B., Waks, J.W. et al. (2025). A comparison of artificial 701 intelligence–enhanced electrocardiography approaches for the prediction of time to mortality 702 using electrocardiogram images. European Heart Journal-Digital Health ...

    work page 2025

Show all 38 references
  1. [9]

    Islam, M.R.U., Tadepalli, P ., and Fern, A. (2025). Self-attention-based diffusion model for 704 time-series imputation in partial blackout scenarios. In Proceedings of the AAAI Conference 705 on Artificial Intelligence vol. 39. pp. 17564–17572. 706

    2025

  2. [10]

    Chen, J., Huang, S., Zhang, Y ., Chang, Q., Zhang, Y ., Li, D., Qiu, J., Hu, L., Peng, X., Du, Y .707 et al. (2024). Congenital heart disease detection by pediatric electrocardiogram based deep 708 learning integrated with human concepts. Nature Communications ��, 976. 709

    2024

  3. [11]

    Wang, J., Qiao, X., Liu, C., Wang, X., Liu, Y ., Y ao, L., and Zhang, H. (2021). Automated ecg710 classification using a non-local convolutional block attention module. Computer methods 711 and programs in biomedicine ��� , 106006. 712

    2021

  4. [12]

    Zhang, S., Lian, C., Xu, B., Su, Y ., and Alhudhaif, A. (2024). 12-lead ecg signal classifica- 713 tion for detecting ecg arrhythmia via an information bottleneck-based multi-scale network. 714 Information Sciences ��� , 120239. 715

    2024

  5. [13]

    Li, J., Aguirre, A.D., Junior, V.M., Jin, J., Liu, C., Zhong, L., Sun, C., Clifford, G., Bran- 716 don Westover, M., and Hong, S. (2025). An electrocardiogram foundation model built on over 717 10 million recordings. NEJM AI �, AIoa2401033. 718

    2025

  6. [14]

    Ji, C., Wang, L., Qin, J., Liu, L., Han, Y ., and Wang, Z. (2024). Msgformer: A multi-scale grid 719 transformer network for 12-lead ecg arrhythmia detection. Biomedical Signal Processing and 720 Control ��, 105499. 721 23

    2024

  7. [15]

    El-Ghaish, H., and Eldele, E. (2024). Ecgtransform: Empowering adaptive ecg arrhythmia 722 classification framework with bidirectional transformer. Biomedical Signal Processing and 723 Control ��, 105714. 724

    2024

  8. [16]

    Jeyarani, A. (2026). An interpretable temporal transformer for 12-lead ecg arrhythmia classi- 725 fication. International Journal of Machine Learning and Cybernetics ��, 14. 726

    2026

  9. [17]

    Yisimitila, T., Wang, C., Hou, M., Maimaitiniyazi, M., Aili, Z., Liu, T., Tan, H., Chellamani, 727 N., Shanmuganathan, M., Liu, M. et al. (2025). Bridging clinical knowledge and ai: an 728 interpretable transformer framework for ecg diagnosis. npj Digital Medicine. 729

    2025

  10. [18]

    Chatterjee, M., Chan, A.D., and Komeili, M. (2026). Toward robust automated cardiovascular 730 arrhythmia detection using self-supervised learning and 1-dimensional vision transformers. 731 Scientific Reports. 732

    2026

  11. [19]

    Y ang, T., Xie, C.X., Huang, H.M., Wang, Y ., Fan, M.H., Kuo, I.C., Chen, T.Y ., Chen, S.L.,733 Chen, C.A., Abu, P .A.R. et al. (2025). Lead analysis for the classification of multi-label 734 cardiovascular diseases and neural network architecture design. Electronics ��, 3211. 735

    2025

  12. [20]

    Dong, J., Wu, H., Zhang, H., Zhang, L., Wang, J., and Long, M. (2023). Simmtm: A simple 736 pre-training framework for masked time-series modeling. Advances in Neural Information 737 Processing Systems ��, 29996–30025. 738

    2023

  13. [21]

    Wang, Y ., Huang, N., Li, T., Y an, Y ., and Zhang, X. (2024). Medformer: A multi-granularity739 patching transformer for medical time-series classification. arXiv preprint arXiv:2405.19363. 740

    2024 arXiv

  14. [22]

    741 (2024)

    Huang, Z., Herbozo Contreras, L.F ., Yu, L., Truong, N.D., Nikpour, A., and Kavehei, O. 741 (2024). S4d-ecg: a shallow state-of-the-art model for cardiac abnormality classification. 742 Cardiovascular Engineering and Technology ��, 305–316. 743

    2024

  15. [23]

    Liu, W., Wang, F ., Huang, Q., Chang, S., Wang, H., and He, J. (2019). Mfb-cbrnn: A 744 hybrid network for mi detection using 12-lead ecgs. IEEE journal of biomedical and health 745 informatics ��, 503–514. 746

    2019

  16. [24]

    Zhu, H., Luo, Q., Wang, Y ., Maimaiti, M., Zhang, H., Xiong, Y ., Ruan, C., Wang, J., Liu, Y .,747 Zhou, M. et al. (2026). Artificial intelligence-enabled 8-channel ecg diagnosing of abnormali- 748 ties with wide qrs complexes: a cohort study. Health Data Science. 749

    2026

  17. [25]

    Y ao, Q., Wang, R., Fan, X., Liu, J., and Li, Y . (2020). Multi-class arrhythmia detection from 12-750 lead varied-length ecg using attention-based time-incremental convolutional neural network. 751 Information Fusion ��, 174–182. 752

    2020

  18. [26]

    Neog, A., Daw, A., Khorasgani, S.F ., and Karpatne, A. (2025). Masking the gaps: An 753 imputation-free approach to time series modeling with missing data. arXiv preprint 754 arXiv:2502.15785. 755

    2025

  19. [27]

    (2024).756 Timexer: Empowering transformers for time series forecasting with exogenous variables

    Wang, Y ., Wu, H., Dong, J., Qin, G., Zhang, H., Liu, Y ., Qiu, Y ., Wang, J., and Long, M. (2024).756 Timexer: Empowering transformers for time series forecasting with exogenous variables. 757 arXiv preprint arXiv:2402.19072. 758

    2024 arXiv

  20. [28]

    Che, Z., Purushotham, S., Cho, K., Sontag, D., and Liu, Y . (2018). Recurrent neural networks 759 for multivariate time series with missing values. Scientific reports �, 6085. 760 24

    2018

  21. [29]

    Chen, X., Li, X., Liu, B., and Li, Z. (2023). Biased temporal convolution graph network for 761 time series forecasting with missing values. In The Twelfth International Conference on 762 Learning Representations. 763

    2023

  22. [30]

    Wang, X., and Y ang, J. (2025). A systematic method combining rotated convolution and state 764 space augmented transformer for digitizing and classifying paper ecgs. Symmetry ��, 120. 765

    2025

  23. [31]

    Holmes: Health online model ensemble serving for deep learning models in intensive 767 care units

    Hong, S., Xu, Y ., Khare, A., Priambada, S., Maher, K., Aljiffry, A., Sun, J., and Tumanov, A.766 (2020). Holmes: Health online model ensemble serving for deep learning models in intensive 767 care units. In Proceedings of the 26th ACM SIGKDD International Conference on Knowle...

    2020

  24. [32]

    Grover, S., Jalali, A., and Etemad, A. (2024). Segment, shuffle, and stitch: A simple layer for 770 improving time-series representations. Advances in Neural Information Processing Systems 771 ��, 4878–4905. 772

    2024

  25. [33]

    Wagner, P ., Strodthoff, N., Bousseljot, R.D., Kreiseler, D., Lunze, F .I., Samek, W., and 773 Schaeffter, T. (2020). Ptb-xl, a large publicly available electrocardiography dataset. Scientific 774 data �, 1–15. 775

    2020

  26. [34]

    Tao, Y ., Zhang, D., Tan, C., Wang, Y ., Shi, L., Chi, H., Geng, S., Ma, Z., Hong, S., and Liu,776 X.P . (2024). An artificial intelligence-enabled electrocardiogram algorithm for the prediction 777 of left atrial low-voltage areas in persistent atrial fibrillation. Journal of...

    2024

  27. [35]

    Tao, Y ., Zhang, D., Pang, N., Geng, S., Tan, C., Tian, Y ., Hong, S., and Liu, X. (2025). 780 Multi-modal artificial intelligence algorithm for the prediction of left atrial low-voltage areas in 781 atrial fibrillation patient based on sinus rhythm electrocardiogram and clini...

    2025

  28. [36]

    Lin, T.Y ., Goyal, P ., Girshick, R., He, K., and Dollár, P . (2017). Focal loss for dense object784 detection. In Proceedings of the IEEE international conference on computer vision. pp. 785 2980–2988. 786

    2017

  29. [37]

    Du, W., Côté, D., and Liu, Y . (2023). Saits: Self-attention-based imputation for time series.787 Expert Systems with Applications ��� , 119619. 788

    2023

  30. [38]

    Nie, Y ., Nguyen, N.H., Sinthong, P ., and Kalagnanam, J. (2022). A time series is worth 64789 words: Long-term forecasting with transformers. arXiv preprint arXiv:2211.14730. 790 Acknowledgements 791 This work was supported by the National Natural Science Foundation of China ...

    2022 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.