Pith. sign in

REVIEW 3 major objections 5 minor 41 references

LV-CadeNet: A Long-View Feature Convolution-Attention Fusion Encoder-Decoder Network for EEG/MEG Spike Analysis

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A convolutional-attention network with six long-view waveform features claims new best results on EEG and MEG spike analysis.

desk verdict The long-view feature idea is genuinely interesting and the experiments are thorough, but the SOTA claim hinges on an underspecified train/test separation in the feature-normalization step that needs to be resolved before the numbers can be trusted. read the letter →

arxiv 2412.08896 v2 pith:5GKU7NMX submitted 2024-12-12 cs.CV

Kuntao Xiao , Xiongfei Wang , Pengfei Teng , Yi Sun , Yong Zhang , Wanli Yang , Zikang Xu , Liang Zhang
show 3 more authors
Hanyang Dong Guoming Luan Shurong Sheng
This is my paper · ORCID
classification cs.CV
keywords EEGspikeclassificationMEGdetectioninterictalepileptiformdischargeslong-viewcontextfeaturesconvolution-attentionencoder-decoderpubliccorpusbalancedaccuracy
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 argues that automated EEG/MEG spike analysis fails not because the models are too weak but because they look only at short windows around a candidate spike. LV-CadeNet adds a long-view channel: six morphological features computed for every complete wave in the signal and normalized against the surrounding roughly six seconds of context, concatenated with the raw segment. On the largest public EEG spike corpus, the full model claims the best balanced accuracy, Cohen's kappa, and weighted F1 among six state-of-the-art baselines; on a clinical MEG dataset, it reports balanced accuracy more than 13 points above the strongest baseline. The paper further claims the long-view features are a plug-and-play module that raises balanced accuracy by more than 8 points across four unrelated architectures.

What carries the argument

The load-bearing object is the seven-channel input matrix $X = [S, F]$, the raw signal segment $S$ concatenated with six long-view feature maps. Each feature map is built from complete waves, where a complete wave is a local maximum between two local minima; the features encode wave topology, amplitude, mean amplitude, left and right slopes, half-width slopes, and sharpness, and each wave's values are z-scored against the surrounding 100 waves, about six seconds of context. CadeNet's encoder stacks convolution-attention blocks in which temporal 1D convolutions and multi-head spatial attention operate on the same representation, with pixel-shuffle downsampling that moves spatial and temporal information into the feature dimension, followed by a decoder that distills the encoder features. The machinery's job is to let the network compare a candidate spike against its long-range background without diluting the input with raw noise.

What would settle it

Re-run the EEG and MEG comparisons with the 100-wave context window constrained never to include the labelled segment's own wave or any test-set time point; if the >8-point balanced-accuracy gains persist, the context-learning claim survives, and if they disappear, the reported gains came from leakage.

Watch

Extended reading notes

Core claim

The central claim is that spike morphology is only meaningful in context: a waveform that looks like a spike in isolation is frequently a non-epileptic transient, and clinicians judge it against a longer stretch of surrounding activity. The paper operationalizes that context as a fixed set of six numeric features per complete wave — topology, amplitude, mean amplitude, slope, half-slope, and sharpness — z-scored over the neighboring 100 waves, and feeds them alongside the raw signal to a convolution-attention encoder-decoder. In the paper's experiments, the combined system reaches 67.32% balanced accuracy, 68.23% Cohen's kappa, and 83.61% weighted F1 on the EEG corpus, and 86.32% balanced accuracy on the clinical MEG test set, compared with 72.74% for the best MEG baseline. The same long-view features added to four unrelated architectures improve their balanced accuracy by more than 8 points, which the paper reads as evidence that the missing ingredient is context, not architecture.

Load-bearing premise

The load-bearing premise is that each segment's roughly six seconds of surrounding context is cleanly available from the same recording without containing the segment's own label or any test-set time points, a guarantee the paper does not spell out.

Editorial extensions

If this is right

  • If the long-view features are as general as claimed, existing EEG/MEG spike detectors can be upgraded by concatenating six normalized feature maps to their input rather than redesigning their architecture.
  • Naive extension of raw input windows hurts all tested models, so contextual information must be supplied as normalized features rather than simply longer clips.
  • The large balanced-accuracy gain on the imbalanced clinical MEG set suggests context features matter most when the class distribution reflects real clinical practice.
  • Ablation results attribute distinct gains to convolutional temporal extraction, progressively increasing spatial/temporal abstraction, and the decoder module, implying each architectural choice matters independently.
  • The 13.58-point MEG balanced-accuracy improvement over the strongest baseline, if it holds, would make automated spike detection substantially more usable in clinical workflows.

Reading between the lines

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

  • Editorial inference: the decisive test for clinical adoption is a strict leave-one-subject-out or recording-level split; the reported gains could shrink if the 100-wave context window crosses file boundaries between training and test sets.
  • Editorial inference: because the features are defined per complete wave, the same construction could be applied to other transient biomarkers such as high-frequency oscillations without changing the network, only the wave detector.
  • Editorial inference: the uniform >8-point improvement across four different architectures is large enough that data leakage through the normalization window is the main alternative explanation to rule out before accepting the context-learning story.
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. The paper proposes LV-CadeNet for EEG/MEG spike classification and detection. The method augments raw signal segments with six hand-crafted "long-view" morphological features, computed at file level and normalized against the surrounding 100 complete waves, and feeds the concatenated input into a new convolution-attention encoder-decoder named CadeNet. On the TUEV EEG dataset, LV-CadeNet-large reports BAcc 67.32, CKap 68.23, and WF1 83.61, slightly exceeding Labram-large, and on a Sanbo MEG test set it reports BAcc 86.32, 13.58 points above the best baseline. Ablation experiments show that adding the long-view features improves SpikeNet, Labram, EEGPT, and SimBA by more than 8 points in balanced accuracy, and architecture ablations attribute gains to convolutional temporal extraction, progressive spatiotemporal abstraction, and the decoder.

Significance. If the central claims hold, the paper makes a useful contribution: it demonstrates a plausible way to inject long-range context into spike analysis without naively extending raw input windows, and the architecture ablations are informative. The strengths include 30 repeated runs with reported standard deviations, evaluation on a public benchmark (TUEV), cross-architecture plug-and-play experiments, and qualitative case inspection. However, the main empirical claim currently rests on a file-level context procedure whose data availability and leakage safety are not specified, and one of the cross-model MEG comparisons is internally inconsistent with the text. The result is defensible but needs clarification and possibly re-analysis before the state-of-the-art conclusion is fully supported.

major comments (3)
  1. [Section 3.2.1 and Section 3.1.1] The load-bearing premise of the long-view features is that every training slice has access to the surrounding 100 complete waves (about 6 seconds) and that the normalization statistics contain no test time points. The paper states that TUEV comprises "1-second bipolar channel segments" and that the official training-test split was used, but it does not state whether these segments are embedded in continuous recordings, whether the official split is file/patient-disjoint or event-level within shared recordings, whether the 100-wave context excludes the wave being normalized, or whether the file-wise z-score of Section 3.1.2 is computed before splitting. These details are decisive because an event-level split within continuous files would let training features use test-set statistics and could explain the large LV gains in Tables 5 and 6. Please clarify these points, and if the current implementation permits overlap, rerun the experiments with strictly disjoint context windows.
  2. [Table 4 and Section 4.2.1] The text says that LV-CadeNet demonstrates "consistent performance superiority over CadeNet" for MEG spike detection, but Table 4 shows that LV-CadeNet is worse than CadeNet on Cohen's Kappa (46.91 vs 51.05), weighted F1 (95.96 vs 96.83), and accuracy. Only balanced accuracy improves. This contradiction weakens the claim that the long-view module is uniformly beneficial. Please report paired significance tests for the differences or explicitly restrict the claim to balanced accuracy.
  3. [Table 3] The state-of-the-art claim over Labram-large rests on differences of 1.16 points in BAcc, 0.78 points in CKap, and 0.32 points in WF1, all of which are within the reported standard deviations (for example, CKap 68.23±3.41 vs 67.45±1.95). Because 30 repeated runs are available, the paper should report paired tests or confidence intervals for these headline comparisons before asserting that LV-CadeNet-large outperforms Labram-large.
minor comments (5)
  1. [Equation (17)] The label smoothing term uses epsilon/2, but the EEG task has N=6 classes and epsilon=0.1; standard label smoothing distributes epsilon/N over the incorrect classes. Please correct the formula or justify the binary-style form.
  2. [Tables 5 and 6] Both tables contain a "Div" column with no entries; either fill it with the reported improvement values or remove the column.
  3. [Section 4.2.2] There is a typo: "TEUV" should be "TUEV".
  4. [Section 4.2.1 and Figure 4] The claim that extending the raw input window degrades performance is described only qualitatively. Please report the numeric results underlying Figure 4 so readers can compare the extended-window baseline with the long-view feature approach.
  5. [Section 3.2.1] The assignment of 0.5/-0.5 to non-extremum points in the topological feature is described verbally; a precise rule (for example, based on proximity to the nearest maximum or minimum) would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the long-view features are deterministic transforms of the raw signal, and reported results are measured on held-out test sets; the TUEV split/context-overlap question is a leakage concern, not a circularity concern.

full rationale

I find no equation-level or definitional circularity. The central new ingredient is the long-view feature construction in Section 3.2.1: F_topo is an extremum encoding of the signal, and F_amp, F_meanAmp, F_slope, F_halfSlope, and F_sharpness are computed from Eqs. (1)-(5) applied to complete waves, then z-scored against a surrounding 100-wave window. These are deterministic transforms of the input signal; they are not defined from the annotation labels, are not fitted to test outcomes, and are not renamed model predictions. The network is trained with cross-entropy on the labeled training slices and evaluated on the predefined TUEV test split and the separate Sanbo-Clinic MEG test collection, so the SOTA claims are externally benchmarked rather than self-referential. The ablation studies also compare architecture variants on the same held-out protocol. The one substantive concern raised in the text is experimental rather than circular: Section 3.2.1 does not state whether the TUEV predefined train/test split is file-disjoint or whether a training slice's 100-wave normalization context can include test time points, so temporal leakage could in principle inflate the reported gains. That is a data-split and control issue, not a case where a prediction reduces to its input by construction, and the review rules require an exhibited reduction to flag circularity. Self-citations such as [28] for the complete-wave definition are not load-bearing because the method is restated in the paper's own equations. The circularity burden is therefore low; I assign 0 rather than a higher score.

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

The central claim rests on the feature construction being a faithful proxy for clinical morphology and on the data pipeline being leak-free. The listed hyperparameters are nonstandard choices that affect the empirical comparison. No new physical or conceptual entities are introduced.

free parameters (4)
  • Long-view normalization context window = 100 complete waves, about 6 seconds
    Chosen based on clinicians' annotation practice rather than derived from data; it defines the central long-view feature and directly affects all results.
  • Attention head count Hslf = 5
    Fixed for all experiments in the self-convolution-attention module; the performance comparison is conditional on this choice.
  • Label smoothing coefficient epsilon = 0.1
    Fixed in Eq. (17) for all models; affects the training objective but is a conventional setting.
  • Learning rate schedule = Initial 5e-4, halved every 5 epochs
    Chosen for all models in the optimization setup; not derived from theory and may influence comparative outcomes.
assumptions (4)
  • domain assumption Wave detection via local extrema correctly decomposes EEG/MEG signals into complete waves.
    Feature construction in Section 3.2.1 relies on reference [28]'s extrema detection; if wave boundaries are wrong, F_topo, F_amp, and related features are distorted.
  • domain assumption The six morphological properties in Eqs. (1) to (5) capture the spike criteria clinicians use.
    Amplitude, slope, half-width slope, sharpness, and mean amplitude are borrowed from reference [19] as proxies for clinical judgment, without independent validation in this paper.
  • domain assumption File-level signals are long enough for the 100-wave, about 6 second, normalization context.
    Step three of Section 3.2.1 assumes surrounding waves exist in the same file; for TUEV clips this is not verified in the manuscript.
  • domain assumption No information leaks from test segments into the file-level features used for training slices.
    Normalization uses the distribution of surrounding waves, and the paper does not specify whether those windows can include test-segment samples or overlapping annotations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LV-CadeNet: A Long-View Feature Convolution-Attention Fusion Encoder-Decoder Network for EEG/MEG Spike Analysis." pith.science (2026). https://pith.science/paper/5GKU7NMX

@misc{pith2026241208896,
  author       = {Pith},
  title        = {Pith review of: LV-CadeNet: A Long-View Feature Convolution-Attention Fusion Encoder-Decoder Network for EEG/MEG Spike Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5GKU7NMX}},
  note         = {Machine review of arXiv:2412.08896}
}
read the original abstract

The analysis of interictal epileptiform discharges (IEDs) in magnetoencephalography (MEG) or electroencephalogram (EEG) recordings represents a critical component in the diagnosis of epilepsy. However, manual analysis of these IEDs, which appear as epileptic spikes, from the large amount of MEG/EEG data is labor intensive and requires high expertise. Although automated methods have been developed to address this challenge, current approaches fail to fully emulate clinical experts' diagnostic intelligence in two key aspects: (1) their analysis on the input signals is limited to short temporal windows matching individual spike durations, missing the extended contextual patterns clinicians use to assess significance; and (2) they fail to adequately capture the dipole patterns with simultaneous positive-negative potential distributions across adjacent sensors that serve as clinicians' key diagnostic criterion for IED identification. To bridge this artificial-human intelligence gap, we propose a novel deep learning framework LV-CadeNet that integrates two key innovations: (1) a Long-View morphological feature representation that mimics expert clinicians' comprehensive assessment of both local spike characteristics and long-view contextual information, and (2) a hierarchical Encoder-Decoder NETwork that employs Convolution-Attention blocks for multi-scale spatiotemporal feature learning with progressive abstraction. Extensive evaluations confirm the superior performance of LV-CadeNet, which outperforms six state-of-the-art methods in EEG spike classification on TUEV, the largest public EEG spike dataset. Additionally, LV-CadeNet attains a significant improvement of 13.58% in balanced accuracy over the leading baseline for MEG spike detection on a clinical MEG dataset from Sanbo Brain Hospital, Capital Medical University.

Figures

Figures reproduced from arXiv: 2412.08896 by the authors.

Figure 1
Figure 1. Framework overview of CadeNet: feature extraction model we propose for spike classification and detection [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. (a) Architecture of our feature extraction model CadeNet when the batch size is 1. (b) Illustration of the [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 2
Figure 2. Subsequently, we aggregate a collection of slices into a batch represented as [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figures from the paper (2 more)
Figure 3
Figure 3. Figure 3: Six exemplary cases where LV-CadeNet correctly identifies non-spike activity while CadeNet fails. Each case [PITH_FULL_IMAGE:figures/full_fig_p011_3.png]
Figure 4
Figure 4. Figure 4: Impact of temporal window extension on model performance using raw signal inputs. [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 39 canonical work pages

  1. [1]

    How common are the “common

    D. Hirtz, D. J. Thurman, K. Gwinn-Hardy, M. Mohamed, A. R. Chaudhuri, and R. Zalutsky, “How common are the “common” neurologic disorders?” Neurology, vol. 68, no. 5, pp. 326–337, 2007. 13

    work page 2007

  2. [2]

    Global, regional, and national burden of epilepsy, 1990–2021: a systematic analysis for the global burden of disease study 2021,

    V . L. Feigin, T. V os, B. S. Nair, S. I. Hay, Y . H. Abate, A. H. Abd Al Magied, S. Abd ElHafeez, A. Abdelkader, M.- A. Abdollahifar, A. Abdullahi et al., “Global, regional, and national burden of epilepsy, 1990–2021: a systematic analysis for the global burden of disease study 2021,” The Lancet Public Health , vol. 10, no. 3, pp. e203–e227, 2025

    work page 1990

  3. [3]

    A review of signal processing and machine learning techniques for interictal epileptiform discharge detection,

    B. Abdi-Sargezeh, S. Shirani, S. Sanei, C. C. Took, O. Geman, G. Alarcon, and A. Valentin, “A review of signal processing and machine learning techniques for interictal epileptiform discharge detection,” Computers in Biology and Medicine, vol. 168, p. 107782, 2024

    work page 2024

  4. [4]

    MEG and EEG in epilepsy,

    G. L. Barkley and C. Baumgartner, “MEG and EEG in epilepsy,” Journal of clinical neurophysiology, vol. 20, no. 3, pp. 163–178, 2003

    work page 2003

  5. [5]

    Development of expert-level automated detection of epileptiform discharges during Electroencephalogram interpretation,

    J. Jing, H. Sun, J. A. Kim, A. Herlopian, I. Karakis, M. Ng, J. J. Halford, D. Maus, F. Chan, M. Dolatshahi, C. Muniz, C. Chu, V . Sacca, J. Pathmanathan, W. Ge, J. Dauwels, A. Lam, A. J. Cole, S. S. Cash, and M. B. West- over, “Development of expert-level automated detection of epileptiform discharges during Electroencephalogram interpretation,” JAMA Neu...

    work page 2020

  6. [6]

    Automatic detection of interictal epileptiform discharges based on time-series sequence merging method,

    J. Zhang, J. Zou, M. Wang, L. Chen, C. Wang, and G. Wang, “Automatic detection of interictal epileptiform discharges based on time-series sequence merging method,” Neurocomputing, vol. 110, pp. 35–43, 2013

    work page 2013

  7. [7]

    Visual and automatic investigation of epileptiform spikes in intracranial EEG recordings,

    M. Dümpelmann and C. Elger, “Visual and automatic investigation of epileptiform spikes in intracranial EEG recordings,” Epilepsia, vol. 40, no. 3, pp. 275–285, 1999

    work page 1999

  8. [8]

    Single-trial classification of MEG recordings,

    M. Perreau Guimaraes, D. K. Wong, E. T. Uy, L. Grosenick, and P. Suppes, “Single-trial classification of MEG recordings,” IEEE Transactions on Biomedical Engineering , vol. 54, no. 3, pp. 436–443, 2007

    work page 2007

Show all 41 references
  1. [9]

    EMS-Net: A deep learning method for autodetecting epileptic Magnetoencephalography spikes,

    L. Zheng, P. Liao, S. Luo, J. Sheng, P. Teng, G. Luan, and J.-H. Gao, “EMS-Net: A deep learning method for autodetecting epileptic Magnetoencephalography spikes,” IEEE Transactions on Medical Imaging , p. 1833–1844, 2020

    2020

  2. [10]

    Satelight: Self-attention-based model for epileptic spike detection from multi-electrode EEG

    K. Fukumori, N. Yoshida, H. Sugano, M. Nakajima, and T. Tanaka, “Satelight: Self-attention-based model for epileptic spike detection from multi-electrode EEG.” Journal of Neural Engineering, p. 055007, 2022

    2022

  3. [11]

    Practical Fundamentals of Clinical MEG Interpretation in Epilepsy,

    C. Laohathai, J. S. Ebersole, J. C. Mosher, A. I. Bagi´c, A. Sumida, G. V on Allmen, and M. E. Funke, “Practical Fundamentals of Clinical MEG Interpretation in Epilepsy,” Frontiers in Neurology, 2021

    2021

  4. [12]

    Criteria for defining interictal epileptiform discharges in EEG,

    M. A. Kural, L. Duez, V . Sejer Hansen, P. G. Larsson, S. Rampp, R. Schulz, H. Tankisi, R. Wennberg, B. M. Bibby, M. Scherg, and S. Beniczky, “Criteria for defining interictal epileptiform discharges in EEG,”Neurology, 2020

    2020

  5. [13]

    V2IED: Dual-view learning framework for detecting events of interictal epileptiform discharges,

    N. Networks, Z. Ming, D. Chen, T. Gao, Y . Tang, W. Tu, and J. Chen, “V2IED: Dual-view learning framework for detecting events of interictal epileptiform discharges,” Neural Networks, vol. 172, p. 106136, 2024

    2024

  6. [14]

    CrossCon- vPyramid: Deep multimodal fusion for epileptic Magnetoencephalography spike detection,

    L. Zhang, S. Sheng, X. Wang, J.-H. Gao, Y . Sun, K. Xiao, W. Yang, P. Teng, G. Luan, and Z. Lv, “CrossCon- vPyramid: Deep multimodal fusion for epileptic Magnetoencephalography spike detection,” IEEE Journal of Biomedical and Health Informatics , vol. 29, no. 5, pp. 3194–3205, 2025

    2025

  7. [15]

    Automated detection of epileptic spikes and seizures incorporating a novel spatial clustering prior,

    H. Dong, S. Sheng, X. Wang, J.-H. Gao, Y . Sun, W. Yang, K. Xiao, P. Teng, G. Luan, and Z. Lv, “Automated detection of epileptic spikes and seizures incorporating a novel spatial clustering prior,” in IEEE International Conference on Bioinformatics and Biomedicine (BIBM) , 202...

    2024

  8. [16]

    Large brain model for learning generic representations with tremendous EEG data in BCI,

    W.-B. Jiang, L.-M. Zhao, and B.-L. Lu, “Large brain model for learning generic representations with tremendous EEG data in BCI,” in The Twelfth International Conference on Learning Representations , 2024

    2024

  9. [17]

    EEGPT: Pretrained transformer for universal and reliable representation of eeg signals,

    G. Wang, W. Liu, Y . He, C. Xu, L. Ma, and H. Li, “EEGPT: Pretrained transformer for universal and reliable representation of eeg signals,” Advances in Neural Information Processing Systems , vol. 37, pp. 39 249–39 280, 2024

    2024

  10. [18]

    Detection of interictal epileptiform discharges using transformer based deep neural network for patients with self-limited epilepsy with centrotemporal spikes,

    P. F. Tong, B. Dong, X. Zeng, L. Chen, and S. X. Chen, “Detection of interictal epileptiform discharges using transformer based deep neural network for patients with self-limited epilepsy with centrotemporal spikes,” Biomedical Signal Processing and Control, vol. 101, p. 107238, 2025

    2025

  11. [19]

    Toward a definition of MEG spike: Parametric description of spikes recorded simultaneously by MEG and depth electrodes,

    R. Nowak, M. Santiuste, and A. Russi, “Toward a definition of MEG spike: Parametric description of spikes recorded simultaneously by MEG and depth electrodes,” Seizure, p. 652–655, Nov 2009

    2009

  12. [20]

    Seizure occurrence and interspike interval: Telemetered Electroen- cephalogram studies,

    J. R. Stevens, B. L. Lonsbury, and S. L. Goel, “Seizure occurrence and interspike interval: Telemetered Electroen- cephalogram studies,” Archives of Neurology, vol. 26, no. 5, pp. 409–419, 1972

    1972

  13. [21]

    Automatic recognition and quantification of interictal epileptic activity in the human scalp EEG

    J. Gotman and P. Gloor, “Automatic recognition and quantification of interictal epileptic activity in the human scalp EEG.” Electroencephalography and clinical neurophysiology, vol. 41 5, pp. 513–29, 1976

    1976

  14. [22]

    Spike detection: Inter-reader agreement and a statistical turing test on a large data set,

    M. L. Scheuer, A. Bagic, and S. B. Wilson, “Spike detection: Inter-reader agreement and a statistical turing test on a large data set,” Clinical Neurophysiology, vol. 128, no. 1, p. 243–250, 2016. 14

    2016

  15. [23]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren et al., “Deep residual learning for image recognition,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016

    2016

  16. [24]

    BIOT: Biosignal transformer for cross-data learning in the wild,

    C. Yang, M. B. Westover, and J. Sun, “BIOT: Biosignal transformer for cross-data learning in the wild,” in Thirty-seventh Conference on Neural Information Processing Systems , 2023

    2023

  17. [25]

    EEG2Rep: enhancing self-supervised EEG representation through informative masked inputs,

    N. Mohammadi Foumani, G. Mackellar, S. Ghane, S. Irtza, N. Nguyen, and M. Salehi, “EEG2Rep: enhancing self-supervised EEG representation through informative masked inputs,” inProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2024, pp. 5544–5555

    2024

  18. [26]

    The temple university hospital eeg data corpus,

    I. Obeid and J. Picone, “The temple university hospital eeg data corpus,” Frontiers in Neuroscience, 2016

    2016

  19. [27]

    Spatiotemporal signal space separation method for rejecting nearby interference in MEG measurements,

    S. Taulu and J. Simola, “Spatiotemporal signal space separation method for rejecting nearby interference in MEG measurements,” Physics in Medicine & Biology , vol. 51, no. 7, p. 1759, 2006

    2006

  20. [28]

    EMHapp: a pipeline for the automatic detection, localization and visualization of epileptic magnetoencephalographic high-frequency oscillations,

    W. Cui, M. Cao, X. Wang, L. Zheng, Z. Cen, P. Teng, G. Luan, and J.-H. Gao, “EMHapp: a pipeline for the automatic detection, localization and visualization of epileptic magnetoencephalographic high-frequency oscillations,” Journal of Neural Engineering, p. 055009, 2022

    2022

  21. [29]

    Batch normalization: accelerating deep network training by reducing internal covariate shift,

    S. Ioffe and C. Szegedy, “Batch normalization: accelerating deep network training by reducing internal covariate shift,” arXiv: Learning, 2015

    2015

  22. [30]

    Gaussian error linear units (GELUs),

    D. Hendrycks and K. Gimpel, “Gaussian error linear units (GELUs),”Cornell University - arXiv,Cornell University - arXiv, 2016

    2016

  23. [31]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” ser. The Neural Information Processing Systems Conference, 2017, p. 6000–6010

    2017

  24. [32]

    Dropout: a simple way to prevent neural networks from overfitting,

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: a simple way to prevent neural networks from overfitting,” Journal of Machine Learning Research,Journal of Machine Learning Research , 2014

    2014

  25. [33]

    Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network,

    W. Shi, J. Caballero, F. Huszár, J. Totz, A. P. Aitken, R. Bishop, D. Rueckert, and Z. Wang, “Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network,” inProceedings of the IEEE conference on computer vision and pattern recog...

    2016

  26. [34]

    Rethinking the Inception architecture for computer vision,

    C. Szegedy, V . Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, “Rethinking the Inception architecture for computer vision,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2016

    2016

  27. [35]

    Energy-guided topology Mamba for EEG-based BCI,

    S. Jeong, J. Jeon, and H.-I. Suk, “Energy-guided topology Mamba for EEG-based BCI,” in 13th International Conference on Brain-Computer Interface (BCI) , 2025, pp. 1–4

    2025

  28. [36]

    Fully-Automated spike detection and dipole analysis of epileptic MEG using deep learning,

    R. Hirano, T. Emura, O. Nakata, T. Nakashima, M. Asai, K. Kagitani-Shimono, H. Kishima, and M. Hirata, “Fully-Automated spike detection and dipole analysis of epileptic MEG using deep learning,” IEEE Transactions on Medical Imaging, p. 2879–2890, 2022

    2022

  29. [37]

    KAN-EEG: towards replacing backbone-MLP for an effective seizure detection system,

    L. F. Herbozo Contreras, J. Cui, L. Yu, Z. Huang, A. Nikpour, and O. Kavehei, “KAN-EEG: towards replacing backbone-MLP for an effective seizure detection system,” Royal Society Open Science , vol. 12, no. 3, p. 240999, 2025

    2025

  30. [38]

    Simba: Simplified mamba-based architecture for vision and multivariate time series,

    B. N. Patro and V . S. Agneeswaran, “Simba: Simplified mamba-based architecture for vision and multivariate time series,” arXiv preprint arXiv:2403.15360, 2024

    2024 arXiv

  31. [39]

    KAN: Kol- mogorov–arnold networks,

    Z. Liu, Y . Wang, S. Vaidya, F. Ruehle, J. Halverson, M. Soljacic, T. Y . Hou, and M. Tegmark, “KAN: Kol- mogorov–arnold networks,” in The Thirteenth International Conference on Learning Representations , 2025

    2025

  32. [40]

    Mamba: Linear-time sequence modeling with selective state spaces,

    A. Gu and T. Dao, “Mamba: Linear-time sequence modeling with selective state spaces,” inFirst Conference on Language Modeling, 2024

    2024

  33. [41]

    DeepUNet: A deep fully convolutional network for pixel-level sea-land segmentation,

    R. Li, W. Liu, L. Yang, S. Sun, W. Hu, F. Zhang, and W. Li, “DeepUNet: A deep fully convolutional network for pixel-level sea-land segmentation,” IEEE journal of selected topics in applied earth observations and remote sensing, vol. 11, no. 11, pp. 3954–3962, 2018. 15

    2018

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.