Pith. sign in

REVIEW 4 major objections 6 minor 64 references

Multimodal Forecasting of Sparse Intraoperative Hypotension Events Powered by Language Model

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read IOHFuseLM, a GPT-2-based multimodal framework, predicts intraoperative hypotension 5–15 minutes ahead by fusing patient descriptions with blood-pressure series, and outperforms six forecasting baselines on two real surgical datasets.

desk verdict A competent application paper with code and ablations, but the headline AUC/recall gains are not reproducible until the authors specify how forecasts become event-level scores. read the letter →

arxiv 2505.22116 v3 pith:OQ5HAIL5 submitted 2025-05-28 cs.CL cs.AI

Jintao Zhang , Zirui Liu , Mingyue Cheng , Shilong Zhang , Tingyue Pan , Yitong zhou , Qi Liu , Yanhu Xie This is my paper
classification cs.CLcs.AI
keywords intraoperativehypotensionmultimodallanguagemodeltimeseriesforecastingdiffusiondataaugmentationdomainadaptivepretrainingmeanarterialpressureearlywarningsystemclinicaldecisionsupport
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

IOHFuseLM is a language-model framework for predicting intraoperative hypotension (IOH) — mean arterial pressure staying below 65 mmHg for a continuous minute — from a 15-minute history of MAP readings plus three static patient attributes (age, gender, surgery type). The paper's central claim is that this model, trained in two stages (diffusion-augmented domain-adaptive pretraining, then fine-tuning with an IOH-weighted loss), identifies sparse hypotensive events more accurately than six established forecasting baselines on two real datasets: 74.46% recall and 0.7425 AUC versus 62.68% and 0.7309 for the best baseline at 10-second sampling, with similar margins on the VitalDB cohort. The clinical interest is early warning: the model forecasts the MAP curve 5–15 minutes ahead, so a warning window precedes the confirmatory monitoring window. If the result holds, operating-room monitors could flag at-risk patients minutes before blood pressure actually drops, using only data already collected during surgery.

What carries the argument

The load-bearing mechanism is token-level cross-modal alignment inside the GPT-2 backbone, governed by a patient-specific attention mask. MAP series are cut into fixed-length patches and linearly projected into patch tokens; the clinical description is tokenized with an extended vocabulary; the mask $M_i = \mathbf{1}_{(l+t)/p}\bigl(\mathbf{1}_\eta - m_i\bigr)^{\top}$ zeroes out attention from text tokens to padding positions, so each meaningful token attends only to genuine physiological tokens under a large penalty $\lambda$ (Eq. 10–11). Around this mechanism are three supporting parts: MTRDA, which splits each series into a multi-scale trend (the average of centered sliding-window smoothings) and a residual, then runs a DDPM-style denoiser on the residual to synthesize additional IOH-bearing series; domain-adaptive pretraining, which masks 20% of patch tokens and minimizes MSE on the masked positions over the augmented set; and task fine-tuning with $\mathcal{L} = \mathrm{MSE}_{\mathrm{normal}} + \rho\,\mathrm{MSE}_{\mathrm{IOH}}$, $\rho=10$, which up-weights errors on hypotensive timestamps so the model is biased toward detecting the rare events.

What would settle it

Run the released code on the 10-second Clinical IOH split and recompute recall and AUC using only the stated rule — more than 60% of forecast MAP values below 65 mmHg within a one-minute window — and check whether a single risk score per instance underlies the AUC; if the headline numbers (74.46% recall, 0.7425 AUC) change materially under the literal rule, or turn out to be computed pointwise on hypotensive timestamps rather than from forecast events, the claimed superiority over GPT4TS (62.68%, 0.7309) is not yet demonstrated.

Watch

Extended reading notes

Core claim

The discovery the paper claims is that the scarcity and heterogeneity of IOH events — not the forecasting architecture — is the binding constraint, and that a multimodal language model breaks it. Static attributes are turned into structured clinical narratives by GPT-4o (PCDG), sparse MAP series are decomposed into trend plus residual and enriched by a diffusion generator (MTRDA), and the two modalities are fused at token level under a patient-specific attention mask before a GPT-2 backbone forecasts the future MAP series. On the Clinical IOH dataset (1,452 patients) and VitalDB (1,522 recordings), the model reports the best recall and AUC in every setting, with the largest margins exactly where events are sparsest and sampling is coarsest; ablations attribute the gain to pretraining, augmentation, the expanded tokenizer, and the clinical descriptions jointly. The paper frames this as evidence that personalized semantic context plus physiologically realistic augmentation yields clinically usable early warning.

Load-bearing premise

The argument depends on the step that turns a continuous MAP forecast into an event-level score: the paper defines a predicted event as more than 60% of forecast values in a one-minute window below 65 mmHg, but it never specifies how the reported AUC and recall are computed from that rule, and the claimed margins stand or fall on that unspecified scoring protocol.

Editorial extensions

If this is right

  • On the 10-second Clinical IOH split the model reaches 74.46% recall and 0.7425 AUC, against 62.68% and 0.7309 for the best baseline (GPT4TS): roughly eleven more hypotensive events per hundred are caught.
  • Because the forecast horizon is 5–15 minutes and each prediction is split into a two-minute warning window followed by a monitoring window, the output is directly usable as an early-warning alarm rather than a post-hoc label.
  • Every component matters: ablating the clinical descriptions drops recall from 74.46% to 68.62%, removing diffusion augmentation to 67.78%, and removing pretraining entirely to 67.85%.
  • Transfer learning from the Clinical IOH dataset to a new cohort with different window lengths lifts recall from 0.00% to 17.65% and AUC from 0.50 to 0.58, indicating the pretrained representations carry over to unseen settings.
  • The 48 ms inference time on an RTX 4090 meets the real-time responsiveness standard (ISO 80601-2-77:2017) cited in the paper, so the framework is computationally deployable in an operating-room loop.

Reading between the lines

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

  • Editorial extension: since the clinical descriptions are generated from just three attributes (age, gender, surgery type), an ablation that replaces GPT-4o narratives with the raw attribute strings would isolate whether the recall gain comes from the wording or simply from an implicit surgery-type prior on baseline risk.
  • Editorial extension: the same trend–residual diffusion augmentation and weighted fine-tuning recipe transfers directly to other sparse high-risk events the paper itself names — intraoperative hypoxia, ICU sepsis onset, arrhythmia — where the alignment of waveform patches with structured text is equally natural.
  • Editorial extension: fixing the event-scoring protocol in code (the 60%-below-threshold rule applied to continuous forecasts) would let the reported AUROC be re-derived as a true event-level metric and compared against alarm-rate-per-case statistics from the early-warning literature, a comparison the paper does not report.
  • Editorial extension: adding further structured fields to the description — comorbidities, medications, lab values — is a drop-in change at the token level, and the framework's gains on three attributes suggest richer text would likely improve personalization further; this is a direct prediction of the paper's alignment mechanism, not a result it reports.
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

4 major / 6 minor

Summary. The paper proposes IOHFuseLM, a multimodal language-model framework for forecasting intraoperative hypotension (IOH) by predicting future mean arterial pressure (MAP) series from a historical MAP window and static patient attributes encoded as clinical text. The method has two training stages: a diffusion-based augmentation stage (MTRDA) that generates synthetic MAP series from a trend-residual decomposition, followed by domain-adaptive pretraining that aligns text tokens with MAP patch tokens via a patient-specific attention mask, and then task fine-tuning with an IOH-weighted MSE loss. The authors evaluate on a private Clinical IOH dataset and the public VitalDB dataset, reporting MSE, MAE, recall, and AUC, and claim consistent improvement over six baselines. Ablation studies, transfer-learning experiments, qualitative forecasts, and a runtime comparison are also included, and the code is publicly available.

Significance. If the reported results are reproducible and the evaluation is well defined, the work is a useful contribution to clinical decision support: it addresses event sparsity through diffusion augmentation, integrates static and dynamic patient information at the token level, and provides a practical forecasting framework with public code. The ablation study and the transfer experiment to a new cohort are valuable, and the runtime comparison with HMF supports deployability. However, the central claim of 'consistently outperforming' the baselines depends on an incompletely specified event-level evaluation metric, on point estimates without uncertainty quantification, and on a private dataset, so the significance cannot be fully assessed from the manuscript as written.

major comments (4)
  1. [Section 3, 'IOH Event Evaluation'] The definition of a predicted IOH event is given as 'more than 60% of the forecasted MAP values within the same one-minute window fall below this threshold,' but the manuscript never specifies how a continuous risk score is derived from the forecasted MAP series to compute the reported ROC AUC, nor how the reported recall is aggregated from timestamp-level or window-level predictions. Without this mapping, the AUC and recall values in Table 1 are not uniquely defined, and the comparison against baselines is not reproducible. This is load-bearing because the claimed improvement over the strongest baseline (0.7425 vs. 0.7309 AUC, 74.46% vs. 62.68% recall on Clinical IOH at 10 s sampling) is small enough that a different but equally plausible event-scoring rule could change the ranking. Please specify the exact algorithm, including any threshold scanning procedure, and state whether the same conversion is applied to all baselines.
  2. [Table 1 and Appendix B] Table 1 reports only point estimates averaged over three runs, with no standard deviations, confidence intervals, or significance tests for the discriminative metrics. The sensitivity analysis in Appendix D reports uncertainty only for MSE/MAE, not for recall or AUC, so the reader cannot judge whether the differences between IOHFuseLM and GPT4TS or PatchTST are stable. Given the main claim is 'consistently outperforms established baselines,' the authors should report run-level variability for all metrics and, where meaningful, statistical tests or at least per-run ranges.
  3. [Section 3, 'Series Instance Construction'] The statement that 'instances with historical windows overlapping IOH episodes are excluded' introduces a potential selection bias: the model is never trained or evaluated on the clinically important situation in which hypotension is already developing within the historical window, yet that is precisely the regime in which an early-warning system would be used. The paper should either justify this exclusion with a clinical rationale or report an additional evaluation on windows that do contain pre-IOH risk patterns. The adaptive sampling intervals Delta_Normal and Delta_IOH also alter the effective class balance; the paper should specify how evaluation is weighted so that the reported recall and AUC are not artifacts of the sampling scheme.
  4. [Section 4.2, Eq. (9), and Appendix A] The augmented dataset is defined as X2 = X1 union of augmented pairs for all i in [N], where X1 contains all records in the dataset, with no explicit restriction to the training split. If augmentation and domain-adaptive pretraining use validation or test patients, then the subject-independent split is violated and the reported results are optimistically biased. The paper must clarify that augmentation and pretraining are performed exclusively on the training partition, or otherwise demonstrate that no information from held-out patients enters the pretraining stage.
minor comments (6)
  1. [Throughout] There are repeated typos and inconsistent terms: 'domaim' for 'domain' (Sections 1, 4.3, 5.2), 'IOHFuseFM' for 'IOHFuseLM' (Section 1), 'MTSDA' for 'MTRDA' (Figure 3), and 'Broder Impacts' for 'Broader Impacts' (Appendix F).
  2. [Appendix A, Table 4] The VitalDB row of Table 4 is malformed, making the train/validation/test counts unreadable; the table should be reformatted so that each dataset, sampling rate, and prediction horizon is clearly separated.
  3. [Appendix A, Table 5] The dataset label 'CH-OBPB' appears to be a typo for the Clinical IOH dataset, and several rows are visually garbled; please correct the labels and align the hyperparameter entries.
  4. [Section 5.2 and Appendix D] The hyperparameter sensitivity discussion is qualitative and refers to Figure 8 without reporting the numerical values behind the curves; providing the exact scores and ranges would make the sensitivity claims verifiable.
  5. [Appendix G] The Limitations appendix acknowledges sensitivity to data collection protocols and reliance on generated clinical descriptions, but these caveats are not reflected in the abstract or conclusion, which state the performance claims without qualification.
  6. [Section 4.1] The use of GPT-4o to generate clinical descriptions is described as based on physician recommendations and literature, but no examples of the generated text or a validation of its clinical correctness are provided; including sample descriptions and a small manual review would strengthen the reproducibility of the PCDG module.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the IOH prediction pipeline is trained on real and diffusion-augmented data and evaluated against external baselines; no claim reduces to its inputs by construction.

full rationale

The paper's derivation chain is empirical rather than definitional. IOHFuseLM is trained in two stages—domain-adaptive pretraining on diffusion-augmented MAP series and task fine-tuning on real clinical records—and its headline performance is measured against independent baselines (DLinear, PatchTST, Fredformer, HMF, GPT4TS, TimeLLM) on two datasets. No equation defines the reported AUC or recall in terms of a fitted parameter or a self-cited result. The 60% one-minute event rule in Section 3 is a fixed clinical evaluation criterion, not a fitted input, and it is applied consistently to predictions. The only notable self-citation is HMF [12], which is used as a baseline rather than as load-bearing justification for the model's components or evaluation. The unspecified mapping from forecasted MAP values to the ROC/AUC curve is a reproducibility and correctness concern, but it is not a circular dependency: the metric is not defined so that the model's superiority is guaranteed. The paper is self-contained against external benchmarks, and no step reduces to its own inputs.

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

The central claim depends on a few hyperparameters and assumptions about clinical validity and data augmentation. No new physical entities are introduced.

free parameters (5)
  • IOH loss weight ρ = 10
    Weight for MSE on hypotensive timestamps in fine-tuning loss (Eq. 12).
  • Pretraining masking ratio R = 0.2
    Random masking ratio for time-series patch tokens during domain adaptive pretraining.
  • Number of augmented series H = varies (e.g., 3-5)
    Number of diffusion-generated series per original series; listed in Table 5.
  • Sampling intervals ΔNormal and ΔIOH = varies by setting
    Adaptive slicing intervals for negative and positive instances; values in Table 5.
  • Diffusion steps K = 50
    Number of diffusion steps with cosine schedule.
assumptions (4)
  • domain assumption MAP below 65 mmHg for ≥1 minute defines IOH.
    Used to assign labels and evaluate events; clinically established threshold but not universal.
  • domain assumption GPT-4o-generated descriptions capture relevant clinical context.
    The framework assumes the template descriptions (e.g., about hormones and vascular compliance) are informative; no validation that they improve on raw attributes alone.
  • domain assumption Diffusion-augmented series improve representation learning.
    The pretraining on augmented data is expected to boost sensitivity to sparse IOH; ablation supports it but the mechanism is not proven.
  • standard math Standard backpropagation and attention mechanisms.
    The model uses standard neural network operations without formal verification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multimodal Forecasting of Sparse Intraoperative Hypotension Events Powered by Language Model." pith.science (2026). https://pith.science/paper/OQ5HAIL5

@misc{pith2026250522116,
  author       = {Pith},
  title        = {Pith review of: Multimodal Forecasting of Sparse Intraoperative Hypotension Events Powered by Language Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OQ5HAIL5}},
  note         = {Machine review of arXiv:2505.22116}
}
read the original abstract

Intraoperative hypotension (IOH) frequently occurs under general anesthesia and is strongly linked to adverse outcomes such as myocardial injury and increased mortality. Despite its significance, IOH prediction is hindered by event sparsity and the challenge of integrating static and dynamic data across diverse patients. In this paper, we propose \textbf{IOHFuseLM}, a multimodal language model framework. To accurately identify and differentiate sparse hypotensive events, we leverage a two-stage training strategy. The first stage involves domain adaptive pretraining on IOH physiological time series augmented through diffusion methods, thereby enhancing the model sensitivity to patterns associated with hypotension. Subsequently, task fine-tuning is performed on the original clinical dataset to further enhance the ability to distinguish normotensive from hypotensive states. To enable multimodal fusion for each patient, we align structured clinical descriptions with the corresponding physiological time series at the token level. Such alignment enables the model to capture individualized temporal patterns alongside their corresponding clinical semantics. In addition, we convert static patient attributes into structured text to enrich personalized information. Experimental evaluations on two intraoperative datasets demonstrate that IOHFuseLM outperforms established baselines in accurately identifying IOH events, highlighting its applicability in clinical decision support scenarios. Our code is publicly available to promote reproducibility at https://github.com/zjt-gpu/IOHFuseLM.

Figures

Figures reproduced from arXiv: 2505.22116 by the authors.

Figure 1
Figure 1. (a) IOH events are sparse and exhibit substantial inter patient variability in onset time, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Top: Temporal segmentation for IOH prediction. The MAP curve is divided into histor￾ical window (orange), warning window (purple), and monitoring window (red). Bottom: Patients are split by procedure into training, validation, and test sets to ensure subject independence and prevent data leakage. Definition of Intraoperative Hypotension. In￾traoperative hypotension (IOH) is defined accord￾ing to clinically establish… view at source ↗
Figure 3
Figure 3. Illustration of our framework. MTRDA decomposes MAP series into multi-scale, and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of models with and without domaim adaptive pretraining under two [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Comparison of training and inference speed between IOHFuseLM and HMF. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison of MAP prediction results across different models. [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Examples of augmented MAP series of MTRDA under different sampling frequencies. [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Parameter sensitivity analysis on the Clinic IOH dataset. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Illustration of the PCDG Prompt De￾sign framework. To generate patient-specific clinical narratives, we de￾sign a structured prompt that guides the large language model GPT-4o in producing medically grounded de￾scriptions. This prompt incorporates static patient at￾tri…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

64 extracted references · 42 canonical work pages

  1. [1]

    Intraoperative hypotension: Pathophysiology, clinical relevance, and therapeutic approaches

    Karim Kouz, Phillip Hoppe, Luisa Briesenick, and Bernd Saugel. Intraoperative hypotension: Pathophysiology, clinical relevance, and therapeutic approaches. Indian journal of anaesthesia, 64(2):90–96, 2020

    work page 2020

  2. [2]

    Association of intraoperative hypotension and severe postoperative complications during non-cardiac surgery in adult patients: A systematic review and meta- analysis

    Jianghui Cai, Mi Tang, Huaye Wu, Jing Yuan, Hua Liang, Xuan Wu, Shasha Xing, Xiao Yang, and Xiao-Dong Duan. Association of intraoperative hypotension and severe postoperative complications during non-cardiac surgery in adult patients: A systematic review and meta- analysis. Heliyon, 9(5), 2023

    work page 2023

  3. [3]

    Association between intraoperative hypotension and myocardial injury after vascular surgery

    Judith AR Van Waes, Wilton A Van Klei, Duminda N Wijeysundera, Leo Van Wolfswinkel, Thomas F Lindsay, and W Scott Beattie. Association between intraoperative hypotension and myocardial injury after vascular surgery. Survey of anesthesiology, 60(5):212, 2016

    work page 2016

  4. [4]

    Association of intraoperative hypotension with postoperative morbid- ity and mortality: systematic review and meta-analysis

    M Wijnberge, J Schenk, E Bulle, AP Vlaar, K Maheshwari, MW Hollmann, JM Binnekade, BF Geerts, and DP Veelo. Association of intraoperative hypotension with postoperative morbid- ity and mortality: systematic review and meta-analysis. BJS open, 5(1):zraa018, 2021

    work page 2021

  5. [5]

    Intraoperative hypoten- sion and the risk of postoperative adverse outcomes: a systematic review

    EM Wesselink, TH Kappen, HM Torn, AJC Slooter, and W A Van Klei. Intraoperative hypoten- sion and the risk of postoperative adverse outcomes: a systematic review. British journal of anaesthesia, 121(4):706–721, 2018

    work page 2018

  6. [7]

    Machine-learning algorithm to predict hypotension based on high-fidelity arterial pressure waveform analysis

    Feras Hatib, Zhongping Jian, Sai Buddi, Christine Lee, Jos Settels, Karen Sibert, Joseph Rinehart, and Maxime Cannesson. Machine-learning algorithm to predict hypotension based on high-fidelity arterial pressure waveform analysis. Anesthesiology, 129(4):663–674, 2018

    work page 2018

  7. [8]

    Deep learning models for the prediction of intraoperative hypotension

    Solam Lee, Hyung-Chul Lee, Yu Seong Chu, Seung Woo Song, Gyo Jin Ahn, Hunju Lee, Sejung Yang, and Sang Baek Koh. Deep learning models for the prediction of intraoperative hypotension. British journal of anaesthesia, 126(4):808–817, 2021

    work page 2021

  8. [9]

    Prediction of intraoperative hypotension using deep learning models based on non-invasive monitoring devices

    Heejoon Jeong, Donghee Kim, Dong Won Kim, Seungho Baek, Hyung-Chul Lee, Yusung Kim, and Hyun Joo Ahn. Prediction of intraoperative hypotension using deep learning models based on non-invasive monitoring devices. Journal of Clinical Monitoring and Computing, pages 1–9, 2024

    work page 2024

Show all 64 references
  1. [10]

    An algorithm based on deep learning for predicting in-hospital cardiac arrest

    Joon-myoung Kwon, Youngnam Lee, Yeha Lee, Seungwoo Lee, and Jinsik Park. An algorithm based on deep learning for predicting in-hospital cardiac arrest. Journal of the American Heart Association, 7(13):e008678, 2018

    2018

  2. [11]

    Prediction of blood pressure after induction of anesthesia using deep learning: A feasibility study

    Young-Seob Jeong, Ah Reum Kang, Woohyun Jung, So Jeong Lee, Seunghyeon Lee, Misoon Lee, Yang Hoon Chung, Bon Sung Koo, and Sang Hyun Kim. Prediction of blood pressure after induction of anesthesia using deep learning: A feasibility study. Applied Sciences, 9(23):5135, 2019

    2019

  3. [12]

    Hmf: A hybrid multi-factor framework for dynamic intraoperative hypotension prediction

    Mingyue Cheng, Jintao Zhang, Zhiding Liu, Chunli Liu, and Yanhu Xie. Hmf: A hybrid multi-factor framework for dynamic intraoperative hypotension prediction. arXiv preprint arXiv:2409.11064, 2024

    2024

  4. [13]

    A composite multi-attention framework for intraoperative hypotension early warning

    Feng Lu, Wei Li, Zhiqiang Zhou, Cheng Song, Yifei Sun, Yuwei Zhang, Yufei Ren, Xiaofei Liao, Hai Jin, Ailin Luo, et al. A composite multi-attention framework for intraoperative hypotension early warning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume ...

    2023

  5. [14]

    Long short-term memory

    Alex Graves and Alex Graves. Long short-term memory. Supervised sequence labelling with recurrent neural networks, pages 37–45, 2012

    2012

  6. [15]

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

    Yunhao Zhang and Junchi Yan. Crossformer: Transformer utilizing cross-dimension dependency for multivariate time series forecasting. In The eleventh international conference on learning representations, 2023. 10

    2023

  7. [16]

    itransformer: Inverted transformers are effective for time series forecasting

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. itransformer: Inverted transformers are effective for time series forecasting. arXiv preprint arXiv:2310.06625, 2023

    2023 arXiv

  8. [17]

    Timexer: Empowering transformers for time series forecasting with exogenous variables

    Yuxuan Wang, Haixu Wu, Jiaxiang Dong, Guo Qin, Haoran Zhang, Yong Liu, Yunzhong Qiu, Jianmin Wang, and Mingsheng Long. Timexer: Empowering transformers for time series forecasting with exogenous variables. arXiv preprint arXiv:2402.19072, 2024

    2024 arXiv

  9. [18]

    Are transformers effective for time series forecasting? In Proceedings of the AAAI Conference on Artificial Intelligence , volume 37, pages 11121–11128, 2023

    Aixin Zeng, Ming Chen, Lei Zhang, and Qiang Xu. Are transformers effective for time series forecasting? In Proceedings of the AAAI Conference on Artificial Intelligence , volume 37, pages 11121–11128, 2023

    2023

  10. [20]

    Diffusion-ts: Interpretable diffusion for general time series genera- tion

    Xinyu Yuan and Yan Qiao. Diffusion-ts: Interpretable diffusion for general time series genera- tion. arXiv preprint arXiv:2403.01742, 2024

    2024 arXiv

  11. [22]

    Effective data augmentation with diffusion models

    Brandon Trabucco, Kyle Doherty, Max Gurinas, and Ruslan Salakhutdinov. Effective data augmentation with diffusion models. arXiv preprint arXiv:2302.07944, 2023

    2023 arXiv

  12. [23]

    One fits all: Power general time series analysis by pretrained lm

    Tian Zhou, Peisong Niu, Liang Sun, Rong Jin, et al. One fits all: Power general time series analysis by pretrained lm. Advances in neural information processing systems, 36:43322–43355, 2023

    2023

  13. [24]

    Time-llm: Time series forecasting by reprogramming large language models

    Ming Jin, Shiyu Wang, Lintao Ma, Zhixuan Chu, James Y Zhang, Xiaoming Shi, Pin-Yu Chen, Yuxuan Liang, Yuan-Fang Li, Shirui Pan, et al. Time-llm: Time series forecasting by reprogramming large language models. arXiv preprint arXiv:2310.01728, 2023

    arXiv

  14. [25]

    Perioperative hypotension: clinical impact, diagnosis, and therapeutic approaches

    Phillip Hoppe, Karim Kouz, and Bernd Saugel. Perioperative hypotension: clinical impact, diagnosis, and therapeutic approaches. Journal of emergency and critical care medicine , 4, 2020

    2020

  15. [26]

    Early intraoperative hypotension and its associated factors among surgical patients undergoing surgery under general anesthesia: An observational study

    Netsanet Temesgen, Efrem Fenta, Chernet Eshetie, and Moges Gelaw. Early intraoperative hypotension and its associated factors among surgical patients undergoing surgery under general anesthesia: An observational study. Annals of Medicine and Surgery, 71:102835, 2021

    2021

  16. [27]

    Prediction of an acute hypotensive episode during an icu hospitalization with a super learner machine-learning algorithm

    Ményssa Cherifa, Alice Blet, Antoine Chambaz, Etienne Gayat, Matthieu Resche-Rigon, and Romain Pirracchio. Prediction of an acute hypotensive episode during an icu hospitalization with a super learner machine-learning algorithm. Anesthesia & Analgesia, 130(5):1157–1166, 2020

    2020

  17. [28]

    Supervised machine-learning predictive analytics for prediction of postinduction hypotension

    Samir Kendale, Prathamesh Kulkarni, and Rosenberg et al. Supervised machine-learning predictive analytics for prediction of postinduction hypotension. Anesthesiology, 129(4):675– 688, 2018

    2018

  18. [29]

    Predicting hypotension by learning from multivariate mixed responses

    Jodie Ritter, Xiaoyu Chen, Lihui Bai, and Jiapeng Huang. Predicting hypotension by learning from multivariate mixed responses. In Proceedings of the International MultiConference of Engineers and Computer Scientists, 2023

    2023

  19. [30]

    Intraoperative hypotension prediction based on features automatically generated within an interpretable deep learning model

    Eugene Hwang, Yong-Seok Park, Jin-Young Kim, Sung-Hyuk Park, Junetae Kim, and Sung- Hoon Kim. Intraoperative hypotension prediction based on features automatically generated within an interpretable deep learning model. IEEE Transactions on Neural Networks and Learning Systems, 2023

    2023

  20. [31]

    Frequency domain deep learning with non-invasive features for intraoperative hypotension prediction

    Jeong-Hyeon Moon, Garam Lee, Seung Mi Lee, Jiho Ryu, Dokyoon Kim, and Kyung-Ah Sohn. Frequency domain deep learning with non-invasive features for intraoperative hypotension prediction. IEEE Journal of Biomedical and Health Informatics, 2024. 11

    2024

  21. [32]

    A comprehensive survey of time series forecasting: Concepts, challenges, and future directions

    Mingyue Cheng, Zhiding Liu, Xiaoyu Tao, Qi Liu, Jintao Zhang, Tingyue Pan, Shilong Zhang, Panjing He, Xiaohan Zhang, Daoyu Wang, et al. A comprehensive survey of time series forecasting: Concepts, challenges, and future directions. Authorea Preprints, 2025

    2025

  22. [33]

    Ariyo, Adewumi O

    Adebiyi A. Ariyo, Adewumi O. Adewumi, and Charles K. Ayo. Stock price prediction using the arima model. In 2014 UKSim-AMSS 16th International Conference on Computer Modelling and Simulation, pages 106–112, 2014

    2014

  23. [34]

    Convtimenet: A deep hierarchical fully convolutional model for multivariate time series analysis

    Mingyue Cheng, Jiqian Yang, Tingyue Pan, Qi Liu, Zhi Li, and Shijin Wang. Convtimenet: A deep hierarchical fully convolutional model for multivariate time series analysis. In Companion Proceedings of the ACM on Web Conference 2025, pages 171–180, 2025

    2025

  24. [35]

    Gate-variants of gated recurrent unit (gru) neural networks

    Rahul Dey and Fathi M Salem. Gate-variants of gated recurrent unit (gru) neural networks. In 2017 IEEE 60th MWSCAS, pages 1597–1600. IEEE, 2017

    2017

  25. [36]

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

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

    2021

  26. [37]

    Autoformer: Decomposition trans- formers with auto-correlation for long-term series forecasting

    Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. Autoformer: Decomposition trans- formers with auto-correlation for long-term series forecasting. Advances in neural information processing systems, 34:22419–22430, 2021

    2021

  27. [38]

    Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting

    Tian Zhou, Ziqing Ma, Qingsong Wen, Xue Wang, Liang Sun, and Rong Jin. Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting. In International conference on machine learning, pages 27268–27286. PMLR, 2022

    2022

  28. [39]

    Formertime: Hi- erarchical multi-scale representations for multivariate time series classification

    Mingyue Cheng, Qi Liu, Zhiding Liu, Zhi Li, Yucong Luo, and Enhong Chen. Formertime: Hi- erarchical multi-scale representations for multivariate time series classification. In Proceedings of the ACM Web Conference 2023, pages 1437–1445, 2023

    2023

  29. [40]

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

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

    2022 arXiv

  30. [41]

    Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting

    Vijay Ekambaram, Arindam Jati, Nam Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages 459–469, 2023

    2023

  31. [42]

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

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

    2023

  32. [43]

    Cyclenet: enhancing time series forecasting through modeling periodic patterns

    Shengsheng Lin, Weiwei Lin, Xinyi Hu, Wentai Wu, Ruichao Mo, and Haocheng Zhong. Cyclenet: enhancing time series forecasting through modeling periodic patterns. Advances in Neural Information Processing Systems, 37:106315–106345, 2024

    2024

  33. [44]

    Csdi: Conditional score-based diffusion models for probabilistic time series imputation

    Yusuke Tashiro, Jiaming Song, Yang Song, and Stefano Ermon. Csdi: Conditional score-based diffusion models for probabilistic time series imputation. Advances in neural information processing systems, 34:24804–24816, 2021

    2021

  34. [45]

    Non-autoregressive conditional diffusion models for time series prediction

    Lifeng Shen and James Kwok. Non-autoregressive conditional diffusion models for time series prediction. In International Conference on Machine Learning, pages 31016–31029. PMLR, 2023

    2023

  35. [46]

    Retrieval-augmented diffusion models for time series forecasting

    Jingwei Liu, Ling Yang, Hongyan Li, and Shenda Hong. Retrieval-augmented diffusion models for time series forecasting. Advances in Neural Information Processing Systems, 37:2766–2786, 2024

    2024

  36. [47]

    Fdf: Flexible decoupled framework for time series forecasting with conditional denoising and polynomial modeling

    Jintao Zhang, Mingyue Cheng, Xiaoyu Tao, Zhiding Liu, and Daoyu Wang. Fdf: Flexible decoupled framework for time series forecasting with conditional denoising and polynomial modeling. arXiv preprint arXiv:2410.13253, 2024. 12

    2024

  37. [48]

    s2ip-llm: Semantic space informed prompt learning with llm for time series forecasting

    Zijie Pan, Yushan Jiang, Sahil Garg, Anderson Schneider, Yuriy Nevmyvaka, and Dongjin Song. s2ip-llm: Semantic space informed prompt learning with llm for time series forecasting. In Forty-first International Conference on Machine Learning, 2024

    2024

  38. [49]

    Are language models actually useful for time series forecasting? Advances in Neural Information Processing Systems, 37:60162–60191, 2024

    Mingtian Tan, Mike Merrill, Vinayak Gupta, Tim Althoff, and Tom Hartvigsen. Are language models actually useful for time series forecasting? Advances in Neural Information Processing Systems, 37:60162–60191, 2024

    2024

  39. [50]

    Perioperative myocardial injury and the contribution of hypotension

    Daniel I Sessler and Ashish K Khanna. Perioperative myocardial injury and the contribution of hypotension. Intensive care medicine, 44(6):811–822, 2018

    2018

  40. [51]

    Pe- rioperative quality initiative consensus statement on intraoperative blood pressure, risk and outcomes for elective surgery

    Daniel I Sessler, Joshua A Bloomstone, Solomon Aronson, Colin Berry, Tong J Gan, John A Kellum, James Plumb, Monty G Mythen, Michael PW Grocott, Mark R Edwards, et al. Pe- rioperative quality initiative consensus statement on intraoperative blood pressure, risk and outcomes fo...

    2019

  41. [52]

    Formula and nomogram for the sphygmomanometric calculation of the mean arterial pressure

    Eduardo Meaney, Felix Alva, Rafael Moguel, Alejandra Meaney, JUAN ALV A, and RICHARD WEBEL. Formula and nomogram for the sphygmomanometric calculation of the mean arterial pressure. Heart, 84(1):64–64, 2000

    2000

  42. [53]

    The meaning of blood pressure

    S Magder. The meaning of blood pressure. Critical Care, 22(1):257, 2018

    2018

  43. [54]

    Marije Wijnberge, Bart F Geerts, Liselotte Hol, Nikki Lemmers, Marijn P Mulder, Patrick Berge, Jimmy Schenk, Lotte E Terwindt, Markus W Hollmann, Alexander P Vlaar, et al. Effect of a machine learning–derived early warning system for intraoperative hypotension vs standard care...

    2020

  44. [55]

    Language models are unsupervised multitask learners

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9, 2019

    2019

  45. [56]

    Definition of clinically relevant intraoperative hypotension: A data-driven approach

    Mathias Maleczek, Daniel Laxar, Angelika Geroldinger, Andreas Gleiss, Paul Lichtenegger, and Oliver Kimberger. Definition of clinically relevant intraoperative hypotension: A data-driven approach. Plos one, 19(11):e0312966, 2024

    2024

  46. [57]

    Mitigating intraoperative hypotension: a review and update on recent advances

    Wael Saasouh, Navid Manafi, Asifa Manzoor, and George McKelvey. Mitigating intraoperative hypotension: a review and update on recent advances. Advances in Anesthesia, 42(1):67–84, 2024

    2024

  47. [58]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020

    2020

  48. [59]

    Vector quantized diffusion model for text-to-image synthesis

    Shuyang Gu, Dong Chen, Jianmin Bao, Fang Wen, Bo Zhang, Dongdong Chen, Lu Yuan, and Baining Guo. Vector quantized diffusion model for text-to-image synthesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10696–10706, 2022

    2022

  49. [60]

    Adaln: a vision transformer for multidomain learning and predisaster building information extraction from images

    Yunhui Guo, Chaofeng Wang, Stella X Yu, Frank McKenna, and Kincho H Law. Adaln: a vision transformer for multidomain learning and predisaster building information extraction from images. Journal of Computing in Civil Engineering, 36(5):04022024, 2022

    2022

  50. [61]

    A survey on time-series pre-trained models

    Qianli Ma, Zhen Liu, Zhenjing Zheng, Ziyang Huang, Siying Zhu, Zhongzhong Yu, and James T Kwok. A survey on time-series pre-trained models. IEEE Transactions on Knowledge and Data Engineering, 2024

    2024

  51. [62]

    Time-ffm: To- wards lm-empowered federated foundation model for time series forecasting

    Qingxiang Liu, Xu Liu, Chenghao Liu, Qingsong Wen, and Yuxuan Liang. Time-ffm: To- wards lm-empowered federated foundation model for time series forecasting. arXiv preprint arXiv:2405.14252, 2024

    2024 arXiv

  52. [63]

    Intraoperative hypotension–physiologic basis and future directions

    Hamdy Awad, Gabriel Alcodray, Arwa Raza, Racha Boulos, Michael Essandoh, Sujatha Bhandary, and Ryan Dalton. Intraoperative hypotension–physiologic basis and future directions. Journal of Cardiothoracic and Vascular Anesthesia, 36(7):2154–2163, 2022. 13

    2022

  53. [64]

    Fred- former: Frequency debiased transformer for time series forecasting

    Xihao Piao, Zheng Chen, Taichi Murayama, Yasuko Matsubara, and Yasushi Sakurai. Fred- former: Frequency debiased transformer for time series forecasting. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 2400–2410, 2024

    2024

  54. [65]

    Llama: Open and efficient foundation language models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timo- thée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023

    2023 arXiv

  55. [66]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  56. [67]

    Autoregressive denois- ing diffusion models for multivariate probabilistic time series forecasting

    Kashif Rasul, Calvin Seward, Ingmar Schuster, and Roland V ollgraf. Autoregressive denois- ing diffusion models for multivariate probabilistic time series forecasting. In International Conference on Machine Learning, pages 8857–8868. PMLR, 2021. 14 A Dataset Details Table 4: S...

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.