Pith. sign in

REVIEW 5 major objections 4 minor 40 references

AmpliNetECG12: A lightweight SoftMax-based relativistic amplitude amplification architecture for 12 lead ECG classification

T0 review · 5 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Using 280,000 trainable parameters and 50 Hz data, a convolutional network with the new aSoftMax activation and cross-lead kernel sharing reports 80.74% average F1 and 97.0% ROC-AUC on nine-class CPSC2018 arrhythmia classification.

desk verdict Interesting aSoftMax idea, but internal contradictions and a missing protocol make the reported results unattributable. read the letter →

arxiv 2411.13903 v1 pith:FXCNJCQ7 submitted 2024-11-21 eess.SP cs.AIcs.LG

classification eess.SPcs.AIcs.LG
keywords 12-leadECGarrhythmiaclassificationlightweightCNNaSoftMaxactivationkernelweightsharingCPSC2018explainableAIdownsampling
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to show that accurate 12-lead ECG arrhythmia classification does not require large models or high sampling rates. It introduces a convolutional architecture, AmpliNetECG12, built around a new activation function called aSoftMax, defined as $A(x) = x \odot \operatorname{softmax}(|x|)$, which amplifies the amplitude of ECG deflections in a normalized way. The model shares convolutional kernels across the 12 leads, treating leads as separate spatial dimensions rather than channels, which cuts trainable parameters to 280,000. On the CPSC2018 dataset, with signals downsampled to 50 Hz, the paper reports an average F1 of 80.74%, a ROC-AUC of 97.0%, and an accuracy of about 84%, outperforming several lightweight baselines at a fraction of their compute and storage. If these numbers hold, the architecture makes portable, resource-limited arrhythmia screening more practical.

What carries the argument

The machinery has three parts. The first is aSoftMax, an activation defined by $A(x) = x \odot \operatorname{softmax}(|x|)$; by applying softmax to absolute values and multiplying back by the original signal, it acts as a relativistic amplifier that boosts prominent ECG deflections while suppressing weak ones, and it is used after every convolutional block. The second is kernel weight sharing across leads: ECG leads are stacked as separate dimensions rather than channels, and the same 1D kernels are convolved over each lead, so the network learns features common to all leads instead of training one extractor per lead. The third is a compact body of four convolutional blocks with layer normalization, lead-wise min and max pooling whose outputs are concatenated, followed by a SoftMax classifier; training uses Adamax and categorical focal loss with label smoothing.

What would settle it

Re-run the model on CPSC2018 with a fully documented stratified split and identical 50 Hz preprocessing, and compare per-class F1 with LightX3ECG and the other baselines; if the average F1 is not close to 80.7% or does not beat them, the efficiency claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a deliberately small CNN can match or beat larger lightweight models on nine-class arrhythmia classification when two design choices are combined: aSoftMax activation, which computes $A(x) = x \odot \operatorname{softmax}(|x|)$ so that the sign of each sample is preserved while its magnitude is reweighted relative to the whole tensor, and kernel weight sharing across leads, which uses the same filters for every lead and thereby captures features that are consistent across the 12-lead layout. On CPSC2018 the model reports average F1 80.74%, ROC-AUC 97.0%, precision 81.90%, recall 80.33%, with 280,000 trainable parameters and a disk footprint of 1.01 MB. The authors interpret the result as evidence that compressed 50 Hz data retains enough morphology for reliable detection and that the stochastic character of aSoftMax improves interpretability by highlighting the ECG segments that drive each arrhythmia decision.

Load-bearing premise

The load-bearing premise is that the model's reported F1 and AUC reflect the same train/test split and preprocessing as the baselines, but the paper never states the split ratio, number of training epochs, batch size, or random seed.

Editorial extensions

If this is right

  • A 280,000-parameter model that works on 50 Hz data can run on wearable and portable devices with limited memory, battery, and transmission bandwidth.
  • Downsampling 12-lead ECGs by a factor of 10 reduces storage and data-transfer costs without, on the reported numbers, sacrificing classification quality.
  • Shared kernels across leads imply that a single feature extractor can serve all leads, which should improve generalization across lead layouts and reduce overfitting on lead-specific noise.
  • The aSoftMax activation's amplification of deflections gives a built-in explanation channel: the same weights that classify also indicate which parts of the ECG waveform mattered, supporting clinician-facing interpretability.
  • The model establishes a new lightweight operating point on CPSC2018, at a fraction of the parameters of the lightweight baselines it is compared with.

Reading between the lines

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

  • A test the paper itself does not run is swapping aSoftMax for ReLU, Swish, or plain softmax at equal parameter count; if the F1 gap shrinks, the activation is the active ingredient.
  • Because aSoftMax reweights magnitudes rather than signs, the mechanism could transfer to other multilead physiological signals such as EEG or EMG, where deflection polarity is clinically meaningful.
  • At 50 Hz, the Nyquist limit discards everything above 25 Hz; comparing the same model at 50, 100, and 250 Hz would show where the diagnostic information lives, which the paper leaves open.
  • The interpretability claim could be tested by checking whether the highlighted regions align with clinician-annotated P, QRS, and T waves; the paper does not quantify this alignment.
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 / 4 minor

Summary. The paper proposes AmpliNetECG12, a lightweight convolutional neural network for 12-lead ECG arrhythmia classification on the CPSC2018 dataset. The two claimed innovations are a new activation function called aSoftMax, defined as A(x) = x ⊙ softmax(|x|), and a kernel weight-sharing mechanism applied across ECG leads. The authors report an accuracy of 84%, an F1-score of about 80.7%, and an ROC-AUC of about 96–97% with 280,000 trainable parameters, using 50 Hz downsampled data. The paper also claims interpretability through SHAP values, UMAP visualizations, and feature-map analysis. The central quantitative claims are not verifiable from the manuscript because the architecture is internally inconsistent, the reported metrics are mutually contradictory, and the experimental protocol is under-specified.

Significance. If the reported performance were reproducible, the result would be practically significant: a model with 280k parameters operating on 50 Hz 12-lead ECG that approaches or exceeds lightweight baselines could enable wearable and low-resource deployment. The paper uses a public benchmark, compares against several lightweight baselines, and attempts an explainability analysis. However, the manuscript in its current form does not provide a well-defined architecture, consistent numbers, or a reproducible protocol, so the significance of the claimed contribution cannot be assessed. The absence of any ablation study also leaves the specific role of aSoftMax unverified.

major comments (5)
  1. [§2.3, Table 2] The output shape of the fourth ConvB layer is dimensionally impossible. The preceding ConvB 1x3/s2 layer produces an output of 12 x 750 x 128, and the next ConvB layer is a 1x9 convolution with stride 1. With 'same' padding, a stride-1 convolution preserves the temporal length, so the output cannot become 12 x 1500 x 128; without padding or with explicit upsampling, the spatial dimensions would shrink. As written, the architecture cannot be instantiated, so the model that generated the reported F1/AUC numbers is not defined by the table.
  2. [§4 vs. §2.3, Table 2] Section 4 describes a system that 'processes three input ECG leads independently using three distinct backbones' and introduces a 'Lead-wise Attention module,' neither of which appears in the architecture described in Section 2.3 or in Table 2. These are two different models, and the paper does not state which one produced the experimental results. The headline performance claim therefore cannot be attributed to the proposed architecture with any confidence.
  3. [§1, §3, §6, Table 4] The reported headline metrics are mutually inconsistent. The abstract reports F1 = 80.71% and AUC = 96.00%; Section 3 reports micro-averaged AUC = 0.97, precision 82%, recall 80%, and F1 = 81%; Table 3 gives an average F1 of 0.8074; Table 4 gives F1 = 0.8070; and Section 6 reports F1 = 80.74% and AUC = 97.0%. Likewise, Table 4 reports 0.20B FLOPs and a model size of 1.01 MB, while the text in Section 3.1 states 0.76 billion FLOPs and 1.2 MB. A reader cannot determine the actual claimed performance or computational cost.
  4. [§2.7, Table 4] The experimental protocol is under-specified to the point of non-reproducibility. No split ratio, number of training epochs, batch size, random seed, or exact preprocessing pipeline is given. The statement that records were 'stratified and split' does not specify the fraction or the stratification variable. Additionally, the paper describes CPSC2018 as multi-label but then says labels were 'one-hot encoded to transform the problem into a multi-class classification task'; if the dataset contains records with more than one label, this transformation is not well defined without further explanation. These omissions make the comparison in Table 4 unverifiable.
  5. [§2.5] The central novelty, aSoftMax, is not isolated or validated. The activation is written as A(x) = x ⊙ softmax(|x|), but the text refers to it as 'Eq. X' and the promised plot and first-order derivative are absent. More importantly, no ablation replaces aSoftMax with ReLU, Swish, or standard softmax under the same architecture; without such an experiment, the paper provides no evidence that the activation contributes to the reported results. The claim in the abstract that aSoftMax has 'stochastic characteristics' is also unsupported, since the given formula is deterministic.
minor comments (4)
  1. [§3.1] The text cites 'lightX3ECG [46]' and 'ECGNet' without a reference number in the list; the reference list ends at [28], so the comparendum sources are incomplete.
  2. [§2.2] The preprocessing description is unclear about how 1500 samples are selected from recordings of 6–60 seconds: the text says 'excess values been dropped,' but the exact cropping window and alignment procedure are not specified.
  3. [Throughout] There are numerous typos and grammatical errors that impede readability, including 'increasse' in the abstract, 'Comparision' in Table 4, 'ConvB for with Conv2D' in Table 2, and 'captures the summary' mismatched with a plural subject.
  4. [§2.4] The kernel weight-sharing description says leads 'are not stacked as channels but treated as separate dimensions,' but the input tensor is written as 12 x 1500 x 1 and the convolution operations appear to treat the 12 as a channel-like dimension; the relationship between the prose and the tensor shapes is not made explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported CPSC2018 results are empirical external-benchmark evaluations, and no prediction is derived from the activation formula or from self-citation.

full rationale

The paper's central claims—aSoftmax activation A(x)=x⊙softmax(|x|), kernel weight sharing, and the F1/AUC numbers on CPSC2018—are not inputs to each other by construction. The activation function is an explicit closed-form transformation of the input tensor; its evaluation is an independent empirical measurement on a public benchmark with a stratified train/test split. No fitted parameter is renamed as a prediction: no parameter is fit to CPSC2018 and then used to 'predict' a CPSC2018-derived quantity other than the test-set metrics themselves. The paper cites prior work for baselines, optimizers, and the dataset, but no load-bearing premise is justified solely by a self-citation: this is a single-author paper with no prior-work self-citations, and the cited Swish, Adamax, focal loss, and CPSC2018 dataset are external, verifiable references. The superiority of aSoftmax is asserted and empirically reported, not derived from those citations, so there is no uniqueness-imported-from-authors or ansatz-smuggled-in-via-citation pattern. The abstract/conclusion/table numbers differ (80.71/80.74/0.8070; AUC 96.00/97.0) and the architecture description is internally inconsistent, but those issues concern attribution of the empirical result to a well-defined architecture, not circular derivation. Therefore the circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The central empirical claims rest on several hand-picked design choices and domain assumptions. The model's parameter count depends on the chosen filter sizes, kernel shapes, and pooling; the input representation depends on the 10x downsampling and 1500-sample truncation; and the training procedure depends on unspecified hyperparameters such as the focal loss gamma. The paper provides no ablations or sensitivity analyses for any of these choices. The aSoftMax activation is an invented mathematical function with no external validation. The assumptions that ECG morphology is consistent across leads and that 50 Hz data retains diagnostic information are plausible but not tested.

free parameters (6)
  • Downsampling factor to 50 Hz = 10x reduction from 500 Hz
    Chosen in Section 2.2 to compress data; the paper asserts essential ECG details are preserved but provides no systematic comparison of other rates.
  • Segment length of 1500 samples = 1500
    Section 2.2 truncates or pads each lead to 1500 samples; no justification for this length is given.
  • Convolution filter counts = 64, 128
    Table 2 fixes filter numbers at 64 and 128; no search or justification is provided.
  • Initial learning rate = 0.007
    Section 2.7 states an exponential decaying learning rate starting at 0.007; no sensitivity analysis is reported.
  • Label smoothing = 0.3
    Section 2.7 uses categorical focal loss with label smoothing 0.3; no ablation shows the effect of this choice.
  • Focal loss gamma = not reported
    The loss defined in Section 2.6 includes a focusing parameter gamma, but its value is never specified.
assumptions (4)
  • domain assumption ECG morphological features such as QRS complexes and T-waves have a consistent structure across leads, so sharing one kernel across leads is valid.
    Stated in Section 2.4 as the basis for kernel weight sharing.
  • domain assumption Downsampling the CPSC2018 signal from 500 Hz to 50 Hz and cutting each lead to 1500 samples preserves the information needed for all nine diagnostic classes.
    Section 2.2 says the processing 'decreases data size while preserving essential ECG details', but no empirical support is provided for the 50 Hz rate or the 1500-sample length.
  • domain assumption The CPSC2018 labels and the stratified split are correct, and the comparison baselines in Table 4 use the same protocol.
    Section 2.7 describes a stratified split without giving the ratio; Table 4 compares F1 values from other papers without confirming identical train/test splits.
  • ad hoc to paper Softmax of absolute values provides a useful amplitude amplification for ECG signals.
    The activation A(x)=x times softmax(|x|) is introduced in Section 2.5 without a theoretical derivation or a controlled comparison to other activations.
invented entities (1)
  • aSoftMax activation function
    purpose: Amplify or suppress ECG signal amplitudes by weighting each element with the softmax of its absolute value.
    Defined in Section 2.5 as A(x)=x times softmax(|x|). No external validation, no theoretical justification, and no ablation isolating its effect. The name invokes 'relativistic amplitude amplification', but softmax outputs are always between 0 and 1, so the function scales magnitudes down rather than amplifying them in an absolute sense.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AmpliNetECG12: A lightweight SoftMax-based relativistic amplitude amplification architecture for 12 lead ECG classification." pith.science (2026). https://pith.science/paper/FXCNJCQ7

@misc{pith2026241113903,
  author       = {Pith},
  title        = {Pith review of: AmpliNetECG12: A lightweight SoftMax-based relativistic amplitude amplification architecture for 12 lead ECG classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FXCNJCQ7}},
  note         = {Machine review of arXiv:2411.13903}
}
read the original abstract

The urgent need to promptly detect cardiac disorders from 12-lead Electrocardiograms using limited computations is motivated by the heart's fast and complex electrical activity and restricted computational power of portable devices. Timely and precise diagnoses are crucial since delays might significantly impact patient health outcomes. This research presents a novel deep-learning architecture that aims to diagnose heart abnormalities quickly and accurately. We devised a new activation function called aSoftMax, designed to improve the visibility of ECG deflections. The proposed activation function is used with Convolutional Neural Network architecture to includes kernel weight sharing across the ECG's various leads. This innovative method thoroughly generalizes the global 12-lead ECG features and minimizes the model's complexity by decreasing the trainable parameters. aSoftMax, combined with enhanced CNN architecture yielded AmpliNetECG12, we obtain exceptional accuracy of 84% in diagnosing cardiac disorders. AmpliNetECG12 shows outstanding prediction ability when used with the CPSC2018 dataset for arrhythmia classification. The model attains an F1-score of 80.71% and a ROC-AUC score of 96.00%, with 280,000 trainable parameters which signifies the lightweight yet efficient nature of AmpliNetECG12. The stochastic characteristics of aSoftMax, a fundamental element of AmpliNetECG12, improve prediction accuracy and also increasse the model's interpretability. This feature enhances comprehension of important ECG segments in different forms of arrhythmias, establishing a new standard of explainable architecture for cardiac disorder classification.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 37 canonical work pages

  1. [1]

    In the same way, cardiovascular diseases have caused more deaths, from 12.1 million in 1990 to 20.5 million in 2021

    Introduction Heart diseases and strokes have become much more common around the world, almost doubling from 271 million cases in 1990 to 523 million cases in 2019. In the same way, cardiovascular diseases have caused more deaths, from 12.1 million in 1990 to 20.5 million in 2021. CVDs are now becoming more common and killing more people [1]. This has put ...

  2. [2]

    This architectural design showcases improved efficacy in categorizing compressed electrocardiogram (ECG) data

    We propose a new deep learning architecture specifically developed for long-term electrocardiogram (ECG) classification, characterized by fewer trainable parameters. This architectural design showcases improved efficacy in categorizing compressed electrocardiogram (ECG) data

  3. [3]

    aSoftmax activation,

    We propose a novel activation function, termed "aSoftmax activation," designed for transforming electrocardiogram (ECG) signals. This function employs a relativistic amplitude amplification approach to accentuate the peaks in ECG data

  4. [4]

    This performance was achieved with a model comprising only 280,000 trainable parameters and while analyzing ECG data sampled at a frequency of just 50Hz

    Our proposed approach has achieved notable results, attaining an accuracy of 84% and an F1 -score of 0.81 on CPSC2018 dataset . This performance was achieved with a model comprising only 280,000 trainable parameters and while analyzing ECG data sampled at a frequency of just 50Hz

  5. [5]

    Absolute Softmax Activation

    Materials and methods 2.1 Dataset The CPSC -2018 [22], provided as part of the 7th International Conference on Biomedical Engineering and Biotechnology, achieved a noteworthy accomplishment in ECG research by making a complete ECG dataset accessible to the public. This dataset is a valuable resource for the worldwide research community. It consists of 6,8...

  6. [6]

    The model achieved a micro -average AUC of 0

    Results The evaluation of the model's performance in classifying cardiac arrhythmias was thoroughly presented in our study , classification efficacy was provided by the Receiver Operating Characteristic (ROC) curves and the Area Under the Curve (AUC) values, as illustrated in Figure 3. The model achieved a micro -average AUC of 0. 97, precision 82%, recal...

  7. [7]

    Jovic, N

    A. Jovic, N. Bogunovic, Feature Extraction for ECG Time-Series Mining Based on Chaos Theory, 2007 29th International Conference on Information Technology Interfaces, 2007, pp. 63-68

  8. [8]

    Sahoo, B

    S. Sahoo, B. Kanungo, S. Behera, S. Sabut, Multiresolution wavelet transform based feature extraction and ECG classification to detect cardiac abnormalities, Measurement, 108 (2017) 55-66

Show all 40 references
  1. [9]

    Jha, M.H

    C.K. Jha, M.H. Kolekar, Cardiac arrhythmia classification using tunable Q -wavelet transform based features and support vector machine classifier, Biomedical Signal Processing and Control, 59 (2020) 101875

  2. [10]

    Discussion Our system demonstrates competitive performance in diagnosis and interpretation when compared to previous works in ECG signal analysis. The improvements can be attributed to several innovative approaches: (i) We employ a unique strategy of processing three input ECG...

  3. [11]

    This aspe ct of our model underscores its potential for applications where computational resources are limited

    Limitation and future prospects Our approach, which utilizes relativistic amplitude amplification, has demonstrated commendable performance in ECG signal analysis, particularly notable for its efficient use of fewer model parameters while still delivering effective predictions...

  4. [12]

    Conclusion In this research article, we introduce an innovative deep learning architecture that leverages a comprehensive analysis of 12-lead, long-term electrocardiograms (ECGs) in their most condensed form to accurately classify different types of Arrhythmias. Our architectu...

  5. [13]

    G.A. Roth, et al., Global Burden of Cardiovascular Diseases and Risk Factors, 1990 - 2019: Update From the GBD 2019 Study, Journal of the American College of Cardiology, 76 (2020) 2982-3021

  6. [14]

    Maršánová, M

    L. Maršánová, M. Ronzhina, R. Smíšek, M. Vítek, A. Němcová, L. Smital, M. Nováková, ECG features and methods for automatic classification of ventricular premature and ischemic heartbeats: A comprehensive experimental study, Scientific Reports, 7 (2017) 11239

  7. [15]

    Ebrahimi, M

    Z. Ebrahimi, M. Loni, M. Daneshtalab, A. Gharehbaghi, A review on deep learning methods for ECG arrhythmia classification, Expert Systems with Applications: X, 7 (2020) 100033

  8. [16]

    Siontis, P.A

    K.C. Siontis, P.A. Noseworthy, Z.I. Attia, P.A. Friedman, Artificial intelligence -enhanced electrocardiography in cardiovascular disease management, Nature Reviews Cardiology, 18 (2021) 465-478

  9. [17]

    Qibin, Z

    Z. Qibin, Z. Liqing, ECG Feature Extraction and Classification Using Wavelet Transform and Support Vector Machines, 2005 International Conference on Neural Networks and Brain, 2005, pp. 1089-1092

  10. [18]

    Tadejko, W

    P. Tadejko, W. Rakowski, Mathematical Morphology Based ECG Feature Extraction for the Purpose of Heartbeat Classification, 6th International Conference on Computer Information Systems and Industrial Management Applications (CISIM'07), 2007, pp. 322-327

  11. [19]

    Q. Yao, R. Wang, X. Fan, J. Liu, Y. Li, Multi- class Arrhythmia detection from 12 -lead varied - length ECG using Attention -based Time - Incremental Convolutional Neural Network, Information Fusion, 53 (2020) 174-182

  12. [20]

    0.7684 16.61 5.91 58.18 1D-SEResNet34

  13. [21]

    Ramachandran, B

    P. Ramachandran, B. Zoph, Q.V. Le, Searching for Activation Functions, ArXiv, abs/1710.05941 (2018)

  14. [22]

    Bhattacharyya, S

    S. Bhattacharyya, S. Majumder, P. Debnath, M. Chanda, Arrhythmic Heartbeat Classification Using Ensemble of Random Forest and Support Vector Machine Algorithm, IEEE Transactions on Artificial Intelligence, 2 (2021) 260-268

  15. [23]

    Gupta, M

    V. Gupta, M. Mittal, KNN and PCA classifier with Autoregressive modelling during different ECG signal interpretation, Procedia Computer Science, 125 (2018) 18-24

  16. [24]

    Huang, B

    J. Huang, B. Chen, B. Yao, W. He, ECG Arrhythmia Classification Using STFT -Based Spectrogram and Convolutional Neural Network, IEEE Access, 7 (2019) 92871-92880

  17. [25]

    Singh, S.K

    S. Singh, S.K. Pandey, U. Pawar, R.R. Janghel, Classification of ECG Arrhythmia using Recurrent Neural Networks, Procedia Computer Science, 132 (2018) 1290-1297

  18. [26]

    0.7845 16.76 5.91 58.75 TI-ResNet18 [27] 0.7872 11.39 1.42 40.51 InceptionTime

  19. [27]

    Yildirim, U.B

    O. Yildirim, U.B. Baloglu, R. -S. Tan, E.J. Ciaccio, U.R. Acharya, A new approach for arrhythmia classification using deep coded features and LSTM networks, Computer Methods and Programs in Biomedicine, 176 (2019) 121-133

  20. [28]

    Additionally, our model excelled in terms of storage requirements

    0.7352 0.45 2.29 1.63 ECGNet 0.7880 1.03 1.97 3.75 LightX3ECG 0.8004 5.31 1.34 6.52 Ours 0.8070 0.28 0.20 1.01 Our model demonstrated superior performance over these established methods, while having the lowest computational cost, measured in Floating Point Operations Per Seco...

  21. [29]

    R. He, Y. Liu, K. Wang, N. Zhao, Y. Yuan, Q. Li, H. Zhang, Automatic Cardiac Arrhythmia Classification Using Combination of Deep Residual Network and Bidirectional LSTM, IEEE Access, 7 (2019) 102119-102135

  22. [30]

    R. Hu, J. Chen, L. Zhou, A transformer -based deep neural network for arrhythmia detection using continuous ECG signals, Computers in Biology and Medicine, 144 (2022) 105325

  23. [31]

    C. Chen, Z. Hua, R. Zhang, G. Liu, W. Wen, Automated arrhythmia classification based on a combination network of CNN and LSTM, Biomedical Signal Processing and Control, 57 (2020) 101819

  24. [32]

    Ahmad, A

    Z. Ahmad, A. Tabassum, L. Guan, N.M. Khan, ECG Heartbeat Classification Using Multimodal Fusion, IEEE Access, 9 (2021) 100615-100626

  25. [33]

    Zhang, S

    D. Zhang, S. Yang, X. Yuan, P. Zhang, Interpretable deep learning for automatic diagnosis of 12 -lead electrocardiogram, iScience, 24 (2021) 102373

  26. [34]

    E.Y.K. Ng, et al., An Open Access Database for Evaluating the Algorithms of Electrocardiogram Rhythm and Morphology Abnormality Detection, Journal of Medical Imaging and Health Informatics, (2018)

  27. [35]

    Kingma, J

    D.P. Kingma, J. Ba, Adam: A Method for Stochastic Optimization, CoRR, abs/1412.6980 (2014)

  28. [36]

    Ruder, An overview of gradient descent optimization algorithms, ArXiv, abs/1609.04747 (2016)

    S. Ruder, An overview of gradient descent optimization algorithms, ArXiv, abs/1609.04747 (2016)

  29. [37]

    T.-Y. Lin, P. Goyal, R.B. Girshick, K. He, P. Dollár, Focal Loss for Dense Object Detection, 2017 IEEE International Conference on Computer Vision (ICCV), (2017) 2999-3007

  30. [38]

    Zhu, et al., Identification of 27 cardiac abnormalities from multi -lead ECG signals: An ensembled Se -ResNet framework with Sign Loss function, 2020

    Z. Zhu, et al., Identification of 27 cardiac abnormalities from multi -lead ECG signals: An ensembled Se -ResNet framework with Sign Loss function, 2020

  31. [39]

    Q. Yao, R. Wang, X. Fan, J. Liu, Y. Li, Multi- class Arrhythmia detection from 12 -lead varied - length ECG using Attention -based Time - Incremental Convolutional Neural Network, Inf. Fusion, 53 (2020) 174-182

  32. [40]

    Ismail Fawaz, et al., InceptionTime: Finding AlexNet for time series classification, Data Mining and Knowledge Discovery, 34 (2019) 1936 - 1962

    H. Ismail Fawaz, et al., InceptionTime: Finding AlexNet for time series classification, Data Mining and Knowledge Discovery, 34 (2019) 1936 - 1962

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.