Pith. sign in

REVIEW 3 major objections 5 minor 31 references

FADE: Forecasting for Anomaly Detection on ECG

T0 review · 3 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A self-supervised forecaster trained only on normal ECG signals flags anomalies it never saw by comparing its forecast with the real signal.

desk verdict Solid engineering pipeline for ECG anomaly detection, but the evaluation leaks patients between domain adaptation and test, so the headline accuracies are likely optimistic. read the letter →

arxiv 2502.07389 v1 pith:ZFSMF3LZ submitted 2025-02-11 cs.CV

Paula Ruiz-Barroso , Francisco M. Castro , José Miranda , Denisa-Andreea Constantinescu , David Atienza , Nicolás Guil This is my paper · ORCID
classification cs.CV
keywords ECGanomalydetectionself-supervisedlearningdomainadaptationforecastingarrhythmiaSplit-MSElossSlowFastU-Net
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 paper proposes FADE, a deep learning system that detects cardiac anomalies by forecasting the next second of a normal ECG and measuring how far the forecast falls from the real signal. The model is trained only on normal ECG recordings, so it never needs labeled anomalous waveforms; a domain-adaptation step adjusts it to a new sensor or patient population. On the MIT-BIH Arrhythmia database after adaptation, the system reports 83.84% accuracy for anomalous samples and 85.46% for normal ones. The point, if it holds, is that generic anomaly detection can be built from abundant normal data rather than from scarce expert labels of every possible abnormality.

What carries the argument

The load-bearing pieces are three. Split-MSE is a loss that computes MSE separately in an inner band near zero (the segments between R-peaks) and outer bands (the R-peaks themselves), forcing the model to learn both morphology and baseline; the authors report it outperforms plain MSE and all tested variants. NMAE is the distance metric used at test time, combining raw mean absolute error with a normalized difference between label and forecast; it outperforms nine alternative distances in their comparison. Domain adaptation fine-tunes the normal-signal forecaster on normal samples from the target dataset, raising anomaly-detection accuracy by roughly 12 percentage points over the non-adapted model. The architecture itself, two frequency paths from SlowFast feeding a U-Net decoder, is what carries the forecasting, and the ablation shows both paths are needed for balanced accuracy.

What would settle it

Run the detector on sustained arrhythmias, such as continuous atrial fibrillation or ventricular tachycardia with no preceding four seconds of normal rhythm, and measure whether the NMAE threshold still separates them; according to the paper's own discussion, the method should be inapplicable or show a sharp accuracy drop in exactly this setting.

Watch

Extended reading notes

Core claim

The paper's central claim is that anomaly detection in ECG can be reduced to the quality of a forecasting model: a model that accurately predicts the future of normal heart activity will produce a large error exactly where the true signal deviates from normal morphology, and that error separates abnormal beats and rhythms from normal ones. FADE performs this with a SlowFast-style encoder and U-Net-style decoder trained with a morphological loss, then fine-tuned on normal samples from the target domain and paired with a normalized mean-absolute-error distance (NMAE) and a threshold. The result is a binary detector that works without ever having seen an anomalous training sample, detects both beat-level and rhythm-level anomalies, and is evaluated over 25 train/test splits with reported means and standard deviations.

Load-bearing premise

The method assumes that every test sample has a four-second window of normal heart activity immediately before the one-second interval being checked, because the forecast is only meaningful when the input is normal.

Editorial extensions

If this is right

  • Anomaly detection can be trained without labeled anomalous ECG data, using only normal recordings plus a small labeled set for threshold selection.
  • The same forecasting mechanism generalizes to both heartbeat-level abnormalities, such as premature ventricular contractions, and rhythm-level abnormalities, such as ventricular tachycardia and bigeminy.
  • Domain adaptation to a new patient population or sensor setup is a major accuracy lever, improving detection accuracy by about 12 percentage points in the reported experiments.
  • A four-second input window gives the best balance of anomaly and normal accuracy compared with one-, two-, and three-second windows in the temporal-length study.
  • A transformer-based comparison baseline retrained under the same setup scores lower in both forecasting quality and anomaly-detection accuracy, indicating that the forecasting mechanism itself matters.

Reading between the lines

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

  • A natural extension is to use the forecast error itself as a localization signal: the time indices where the distance is largest may mark where in the beat the anomaly begins, something the current binary threshold does not expose.
  • Because the forecaster only needs normal data, the same pipeline could transfer to other physiological signals with sparse abnormal events, such as EEG or photoplethysmography, wherever a reliable normal corpus exists.
  • The four-second normal-window assumption suggests a practical screening design: a wearable could keep a rolling buffer and score only segments whose trailing four seconds pass a normality check, while sustained arrhythmias are handled by a separate trigger.
  • The accuracy gap between rhythm anomalies (97.48%) and beat anomalies (79.78%) implies that morphology-preserving ectopic beats are the harder case, and a targeted loss weighting or longer context might close that gap.
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

3 major / 5 minor

Summary. This paper presents FADE, a self-supervised forecasting model for ECG anomaly detection. The model is trained on normal ECG windows from MIT-BIH NSR using a novel Split-MSE loss, then domain-adapted using normal samples from the MIT-BIH Arrhythmia database, and finally detects anomalies by comparing a forecasted one-second signal with the actual signal through a novel NMAE distance and a threshold. The authors report an average anomaly detection accuracy of 83.84% and normal classification accuracy of 85.46% on MIT-BIH Arrhythmia, and they provide ablation studies, a temporal-length study, and per-anomaly accuracy analysis.

Significance. The paper addresses a relevant problem and proposes an interesting idea: using forecasting of normal ECG as a self-supervised pretext task, avoiding the need for large labeled anomaly datasets. The design of a morphology-aware loss (Split-MSE) and a distance metric (NMAE) is well motivated. The experimental protocol is transparent in many respects (nested train/test splits for threshold selection, 25 combinations, ablation studies) and the authors state that the full code will be released. However, the main accuracy claim currently rests on an evaluation protocol that allows the domain-adaptation training set and the anomaly-detection test set to share subjects, which is a serious threat to the validity of the reported generalization.

major comments (3)
  1. [4.2.3-4.2.4] The domain-adaptation and anomaly-detection evaluation are not patient-disjoint. Section 4.2.3 states that normal samples are 'randomly shuffled and divided according to a ratio of 80% training and 20% test' with no subject-wise separation, and Section 4.2.4 then uses 'all abnormal samples' from the same 35-subject pool in the anomaly test set. Since the domain-adapted model is fine-tuned on normal windows of a subject, its ability to flag an anomalous window from that same subject may reflect subject-specific adaptation rather than a generalizable forecasting model. This is a correctness risk for the central claims of 83.84% anomaly accuracy and 85.46% normal accuracy, and the Discussion does not list patient overlap among the limitations. The experiments should be re-run with patient-disjoint splits (e.g., leaving out entire subjects from domain adaptation when they appear in the anomaly test set) and per-subject results should be reported.
  2. [3.4, 4.2.4] The method is repeatedly described as 'fully unsupervised' (last paragraph of Related Work; Section 6) and 'never having seen anomalous samples,' but the anomaly threshold T is selected using labeled normal and abnormal samples (Section 4.2.4). This is standard calibration rather than fully label-free detection. The claims should be revised to state that the pipeline requires a labeled calibration set of anomalous segments to select the threshold, or the threshold-selection procedure should be changed to a fully unsupervised rule (e.g., a fixed percentile of normal distances).
  3. [5.4] The choice of the input duration WI=4s is made by comparing accuracies on the same test folds that are later used for the reported final performance (Table 2). This is a model-selection-on-test procedure: the reported 84.65% global accuracy is the result of selecting the best of four candidate input lengths on the held-out data. To obtain unbiased estimates, the WI selection should be performed on a separate validation fold, or the paper should present the final accuracy as the result of this selection process with appropriate caveats.
minor comments (5)
  1. [4.2.2] A single Leave-One-Subject-Out evaluation (Patient 5 only) is used to assess the baseline forecasting model; this is not a statistically robust estimate of forecasting performance. Please provide error bars or use multiple LOSO folds.
  2. [5.1] The NMAE comparison between FADE (0.011) and the reimplemented transformer (0.022) is reported without statistical significance or error bars; the conclusion that FADE 'predicts better future signals' is based on this single comparison.
  3. [5.3, Table 3] Table 3 reports per-anomaly accuracies but no sample counts. The text mentions classes with fewer than 12 samples but does not identify them; adding counts would help interpret the 0% accuracy for nodal escape beats.
  4. [Abstract and title] Typographical issues: 'Arrythmia' in the abstract should be 'Arrhythmia', and the title displays 'F ADE' with an unintended space; also 'commit those values for brevity' in Section 5.1 should be 'omit those values'.
  5. [5.6.1] The description of the third Split-MSE variation ('separating positive and negative values into two distinct bands') is underspecified; please define the band limits and how the two bands are combined in the loss.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: forecasting loss, NMAE, and threshold calibration are independent of the reported test accuracies.

full rationale

The reported anomaly-detection accuracies are not forced by construction. The forecasting model is trained self-supervised on normal ECG windows with the Split-MSE loss (Eq. 1), and the NMAE distance (Eq. 2) compares forecast and actual signals. Neither the loss nor the distance is defined in terms of the anomaly labels or the reported accuracy. The anomaly threshold is selected on a subset of held-out normal and abnormal samples, and the final accuracy is computed on the remaining samples (Sections 4.2.4 and 5.3), which is standard calibration rather than circular prediction. The model itself is never trained on anomalous samples, and the paper does not rely on any self-citation chain or imported 'uniqueness' theorem to force its choices. The acknowledged limitations—requiring a normal preceding window and not classifying anomaly type—are scope restrictions, not circular reasoning. A separate evaluation-leakage concern exists: the domain-adaptation normal samples and the anomaly test samples come from the same MIT-BIH Arrhythmia subjects (Sections 4.2.3–4.2.4), so the numbers may overstate cross-patient generalization; however, this is a correctness/generalization issue, not a case where a prediction reduces by definition or by fitting to its own target.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the stationarity of normal ECG over short windows, the assumption that anomalies are preceded by a clean normal segment, and the separability of NMAE distances. All free parameters are derived from or selected on the same datasets used for evaluation, which inflates the reported accuracy somewhat.

free parameters (8)
  • Split-MSE band limits (NSR training) = ±0.4
    Inner band [-0.4, 0.4], outer bands outside; selected via cross-validation empirical setup (Section 4.2.2).
  • Split-MSE weights w1, w2 (NSR training) = w1=5, w2=1
    Balances inner and outer band losses; selected via cross-validation (Section 4.2.2).
  • Split-MSE band limits (domain adaptation) = ±0.3
    Different bounds used when fine-tuning on MIT-BIH Arrhythmia (Section 4.2.3).
  • Split-MSE weights w1, w2 (domain adaptation) = w1=1, w2=1
    Adjusted during fine-tuning (Section 4.2.3).
  • Anomaly threshold T = Not given, searched over percentiles of anomaly distances from 0 to 100
    Chosen to maximize global accuracy on a 20% labeled subset (Section 4.2.4).
  • Input window WI = 4 seconds
    Selected by comparing 1, 2, 3, and 4 second inputs; 4 seconds gives best accuracy (Section 5.4).
  • Signal quality threshold = 0.5 (50%)
    Used to remove noisy NSR samples (Section 4.2.1).
  • Amplitude clipping limits = [-0.75, 2.5]
    Chosen from histogram of max/min values in the NSR dataset (Section 4.2.1).
assumptions (4)
  • domain assumption The next one second of a normal ECG is predictable from the preceding four seconds under the given filters.
    The whole forecasting approach relies on this stationarity. No physiological argument beyond the choice WL=1 beat is given (Section 4.2.1).
  • domain assumption Anomalies of interest occur within the one-second label window and are preceded by a four-second normal window.
    Subject selection criterion (iii) and the slicing in Figure 2; if anomalies overlap the input window, detection breaks.
  • domain assumption NMAE distance separates normal from anomalous segments well enough for a single global threshold.
    Threshold search assumes a bimodal distance distribution; this is validated empirically only on this dataset.
  • domain assumption Normal signals from MIT-BIH NSR and MIT-BIH Arrhythmia are drawn from the same distribution after resampling and filtering.
    Domain adaptation via fine-tuning assumes transferability; results support it but it is not proven generally.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FADE: Forecasting for Anomaly Detection on ECG." pith.science (2026). https://pith.science/paper/ZFSMF3LZ

@misc{pith2026250207389,
  author       = {Pith},
  title        = {Pith review of: FADE: Forecasting for Anomaly Detection on ECG},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZFSMF3LZ}},
  note         = {Machine review of arXiv:2502.07389}
}
read the original abstract

Cardiovascular diseases, a leading cause of noncommunicable disease-related deaths, require early and accurate detection to improve patient outcomes. Taking advantage of advances in machine learning and deep learning, multiple approaches have been proposed in the literature to address the challenge of detecting ECG anomalies. Typically, these methods are based on the manual interpretation of ECG signals, which is time consuming and depends on the expertise of healthcare professionals. The objective of this work is to propose a deep learning system, FADE, designed for normal ECG forecasting and anomaly detection, which reduces the need for extensive labeled datasets and manual interpretation. FADE has been trained in a self-supervised manner with a novel morphological inspired loss function. Unlike conventional models that learn from labeled anomalous ECG waveforms, our approach predicts the future of normal ECG signals, thus avoiding the need for extensive labeled datasets. Using a novel distance function to compare forecasted ECG signals with actual sensor data, our method effectively identifies cardiac anomalies. Additionally, this approach can be adapted to new contexts through domain adaptation techniques. To evaluate our proposal, we performed a set of experiments using two publicly available datasets: MIT-BIH NSR and MIT-BIH Arrythmia. The results demonstrate that our system achieves an average accuracy of 83.84% in anomaly detection, while correctly classifying normal ECG signals with an accuracy of 85.46%. Our proposed approach exhibited superior performance in the early detection of cardiac anomalies in ECG signals, surpassing previous methods that predominantly identify a limited range of anomalies. FADE effectively detects both abnormal heartbeats and arrhythmias, offering significant advantages in healthcare through cost reduction or processing of large-scale ECG data.

Figures

Figures reproduced from arXiv: 2502.07389 by the authors.

Figure 1
Figure 1. Proposed System. General functionality of our proposed system for ECG forecasting with domain adaptation and heart anomaly detection. 3. Methods A sketch of our proposed system is shown in [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. ECG signal slicing. ECG slices for training the forecasting model: I represents the input data, L is the label data and AL shows the auxiliary label data. 8 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Proposed Model. The proposed model for ECG forecasting that combines both SlowFast and U-Net networks is shown. The U-Net based decoder is on the left side (blue). The Slow path can be observed in the middle (color pink), while the Fast path is on the right side (green). Note that orange arrows indicate concatenation operation and black arrows show connections between layers. More details in the main text. the model… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Split-MSE Loss Function. MSE is calculated separately in two different ranges: outer bands, designed to capture the R-peaks, and inner band, which seeks to seize the segment between R-peaks. Note that P represents P-wave (atrial depolarization), QRS complex represents …
Figure 5
Figure 5. Figure 5: Domain Adaptation. For a given sample of a new domain, the top row shows the prediction (unseen temporal segment) using the original model, while the bottom row represents the prediction employing the domain-adapted model. Huge differences can be appreciated in the uns…
Figure 6
Figure 6. Figure 6: Example of an ECG forecasting comparison for the same input. (a) ECG forecasting trained with MSE. (b) ECG forecasting trained with Split-MSE. Vertical dash-green lines separate the previously seen signal in the input and the new second unseen by the model. Purple dott…
Figure 7
Figure 7. Figure 7: Examples of ECG forecasting for different patients. The top row shows normal signals, while the bottom row shows anomalous signals. Note that blue represents the real ECG and red represents the predicted signal. 5.3. Heart Anomalies Detection Two types of experiments a…
Figure 8
Figure 8. Figure 8: c represents the combination of the heat maps of both paths. We can conclude that the model focuses mostly on R-peaks: Fast Path exactly on the R-peaks and Slow Path on their surroundings, whereas the rest of the ECG has more or less the same importance for the model. …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 27 canonical work pages

  1. [1]

    Anbalagan, M

    T. Anbalagan, M. K. Nath, D. Vijayalakshmi, A. Anbalagan, Analysis of various techniques for ECG signal in healthcare, past, present, and future, Biomedical Engineering Advances 6 (2023) 100089. doi:10.1016/j.bea.2023.100089. URL https://www.sciencedirect.com/science/article/pii/ S2667099223000191

    work page arXiv 2023

  2. [2]

    Nezamabadi, N

    K. Nezamabadi, N. Sardaripour, B. Haghi, M. Forouzanfar, Unsuper- vised ecg analysis: A review, IEEE Reviews in Biomedical Engineering 16 (2023) 208–224. doi:10.1109/RBME.2022.3154893

    work page arXiv 2023

  3. [3]

    Anomaly Detection in Electrocardiograms: Advancing Clinical Diagnosis Through Self-Supervised Learning

    A. Jiang, C. Huang, Q. Cao, Y. Xu, Z. Zeng, K. Chen, Y. Zhang, Y. Wang, Anomaly detection in electrocardiograms: Advancing clinical diagnosis through self-supervised learning, arXiv preprint arXiv:2404.04935 (2024)

    work page Pith review arXiv 2024

  4. [4]

    A. L. Goldberger, L. A. Amaral, L. Glass, J. M. Hausdorff, P. C. Ivanov, R. G. Mark, J. E. Mietus, G. B. Moody, C.-K. Peng, H. E. Stanley, Phys- iobank, physiotoolkit, and physionet: components of a new research re- source for complex physiologic signals, circulation 101 (23) (2000) e215– e220

    work page 2000

  5. [5]

    G. B. Moody, R. G. Mark, The impact of the mit-bih arrhythmia database, IEEE engineering in medicine and biology magazine 20 (3) (2001) 45–50. 30

    work page 2001

  6. [6]

    S. Tan, G. Androz, A. Chamseddine, P. Fecteau, A. Courville, Y. Bengio, J. P. Cohen, Icentia11k: An unsupervised representa- tion learning dataset for arrhythmia subtype discovery, arXiv preprint arXiv:1910.09570 (2019)

    arXiv 2019

  7. [7]

    Z. Liu, A. Alavi, M. Li, X. Zhang, Self-supervised contrastive learning for medical time series: A systematic review, Sensors 23 (9) (2023) 4221

    work page 2023

  8. [8]

    J. Lai, H. Tan, J. Wang, L. Ji, J. Guo, B. Han, Y. Shi, Q. Feng, W. Yang, Practical intelligent diagnostic algorithm for wearable 12-lead ecg via self-supervised learning on large-scale dataset, Nature Communications 14 (1) (2023) 3741

    2023

Show all 31 references
  1. [9]

    Li, et al., Self-supervised contrastive learning for ecg anomaly de- tection, Journal of Biomedical Informatics 113 (2021) 103759

    X. Li, et al., Self-supervised contrastive learning for ecg anomaly de- tection, Journal of Biomedical Informatics 113 (2021) 103759. doi: 10.1016/j.jbi.2020.103759

    2021

  2. [10]

    F. D. Pup, M. Atzori, Applications of self-supervised learning to biomed- ical signals: A survey, IEEE Access 11 (2023) 144180–144203. doi: 10.1109/ACCESS.2023.3344531

    2023

  3. [11]

    Kiyasseh, T

    D. Kiyasseh, T. Zhu, D. A. Clifton, Clocs: Contrastive learning of car- diac signals across space, time, and patients, in: International Confer- ence on Machine Learning, PMLR, 2021, pp. 5606–5615

    2021

  4. [12]

    C. T. Wei, M.-E. Hsieh, C.-L. Liu, V. S. Tseng, Contrastive heart- beats: Contrastive learning for self-supervised ecg representation and phenotyping, in: ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2022, pp. 1126–1130

    2022

  5. [13]

    X. Lan, D. Ng, S. Hong, M. Feng, Intra-inter subject self-supervised learning for multivariate cardiac signals, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 36, 2022, pp. 4532–4540

    2022

  6. [14]

    T. Phan, D. Le, P. Brijesh, D. Adjeroh, J. Wu, M. O. Jensen, N. Le, Mul- timodality multi-lead ecg arrhythmia classification using self-supervised learning, in: 2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI), IEEE, 2022, pp. 01–04. 31

    2022

  7. [15]

    Y. Jin, Z. Li, Y. Tian, X. Wei, C. Liu, A self-supervised framework for computer-aided arrhythmia diagnosis, Applied Soft Computing (2024) 112024

    2024

  8. [16]

    Zheng, Z

    Y. Zheng, Z. Liu, R. Mo, Z. Chen, W.-s. Zheng, R. Wang, Task-oriented self-supervised learning for anomaly detection in electroencephalogra- phy, in: International Conference on Medical Image Computing and Computer-Assisted Intervention, Springer, 2022, pp. 193–203

    2022

  9. [17]

    Mehari, N

    T. Mehari, N. Strodthoff, Self-supervised representation learning from 12-lead ecg data, Computers in biology and medicine 141 (2022) 105114

    2022

  10. [18]

    C. Luo, G. Wang, Z. Ding, H. Chen, F. Yang, Segment origin predic- tion: A self-supervised learning method for electrocardiogram arrhyth- mia classification, in: 2021 43rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2021, pp....

    2021

  11. [19]

    B. T. Lee, S. T. Kong, Y. Song, Y. Lee, Self-supervised learning with electrocardiogram delineation for arrhythmia detection, in: 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), IEEE, 2021, pp. 591–594

    2021

  12. [20]

    Zhang, S

    W. Zhang, S. Geng, S. Hong, A simple self-supervised ecg represen- tation learning method via manipulated temporal–spatial reverse de- tection, Biomedical Signal Processing and Control 79 (2023) 104194. doi:https://doi.org/10.1016/j.bspc.2022.104194

    2023

  13. [21]

    Alamr, A

    A. Alamr, A. Artoli, Unsupervised transformer-based anomaly detection in ecg signals, Algorithms 16 (3) (2023) 152. doi:10.3390/a16030152

    2023 doi

  14. [22]

    Dutta, R

    K. Dutta, R. Lenka, S. R. Nayak, A. Khandual, A. K. Bhoi, Med-net: a novel approach to ecg anomaly detection using lstm auto-encoders, International Journal of Computer Applications in Technology 65 (4) (2021) 343–357

    2021

  15. [23]

    T. Kite, U. T. Siam, B. Ayers, N. Houstis, A. D. Aguirre, Unlock- ing telemetry potential: Self-supervised learning for continuous clinical electrocardiogram monitoring, arXiv preprint arXiv:2406.16915 (2024). 32

    2024 arXiv

  16. [24]

    Carrera, B

    D. Carrera, B. Rossi, P. Fragneto, G. Boracchi, Online anomaly detection for long-term ecg monitoring using wearable de- vices, Pattern Recognition 88 (2019) 482–492. doi:https: //doi.org/10.1016/j.patcog.2018.11.019. URL https://www.sciencedirect.com/science/article/pii/ S003...

    2019 doi

  17. [25]

    Zhang, J

    Y. Zhang, J. Wang, Y. Chen, H. Yu, T. Qin, Adaptive memory networks with self-supervised learning for unsupervised anomaly detection, IEEE Transactions on Knowledge and Data Engineering 35 (12) (2023) 12068– 12080. doi:10.1109/TKDE.2021.3139916

    2023

  18. [26]

    Feichtenhofer, H

    C. Feichtenhofer, H. Fan, J. Malik, K. He, Slowfast networks for video recognition, in: Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 6202–6211

    2019

  19. [27]

    Ronneberger, P

    O. Ronneberger, P. Fischer, T. Brox, U-net: Convolutional networks for biomedical image segmentation, in: Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international con- ference, Munich, Germany, October 5-9, 2015, proceedings, part III 18, Spri...

    2015

  20. [28]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778

    2016

  21. [29]

    Ostchega, K

    Y. Ostchega, K. S. Porter, J. Hughes, C. F. Dillon, T. Nwankwo, Resting pulse rate reference data for children, adolescents, and adults; united states, 1999-2008 (2011)

    2011

  22. [30]

    Zhang, X

    D. Zhang, X. Shen, X. Qi, Resting heart rate and all-cause and car- diovascular mortality in the general population: a meta-analysis, Cmaj 188 (3) (2016) E53–E63

    2016

  23. [31]

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, D. Batra, Grad-cam: Visual explanations from deep networks via gradient-based localization, in: Proceedings of the IEEE international conference on computer vision, 2017, pp. 618–626. 33

    2017

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.