Pith. sign in

REVIEW 3 major objections 5 minor 7 cited by

This paper aims to establish that EEG foundation models, despite rapid progress, have not yet achieved universally transferable representations: across a 13-dataset benchmark, specialist models trained from scratch remain highly competitive

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 06:14 UTC pith:IB2KSPA7

load-bearing objection A genuinely useful EEG foundation-model benchmark and survey, but the 'bigger is not better' conclusion is a confounded cross-model observation, not a scaling-law test. the 3 major comments →

arxiv 2601.17883 v3 pith:IB2KSPA7 submitted 2026-01-25 cs.LG cs.CV

EEG-FM-Compass: Progress, Benchmarking, and Future Directions for EEG Foundation Models

Dingkun Liu , Yuheng Chen , Zhu Chen , Zhenyao Cui , Yaozhi Wen , Jiayu An , Jingwei Luo , Dongrui Wu This is my paper
classification cs.LG cs.CV
keywords EEG foundation modelsbrain-computer interfacebenchmarkself-supervised pre-trainingtransfer learningmodel scalinglinear probingfew-shot calibration
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tries to establish that EEG foundation models, despite rapid progress, have not yet delivered broadly transferable representations. It builds a benchmark of 12 open-source foundation models versus 7 specialist baselines across 13 datasets and nine BCI paradigms, under both cross-subject and within-subject few-shot settings. Its central findings are that frozen encoders (linear probing) often underperform, specialist models trained from scratch remain highly competitive, and larger model scale does not reliably improve generalization. If these results hold, the field's emphasis on scaling up pre-training should shift toward data quality, paradigm-specific pre-training, and efficient adaptation.

Core claim

The paper's central claim is that current EEG foundation models do not yet justify the foundation-model premise: they rarely outperform small specialist models trained from scratch, and they cannot be used as frozen feature extractors. The aggregate ranking places CBraMod (4.0M parameters) first and EEGNet (2K parameters) second, with four of the top five models being specialists. The authors interpret this as evidence that masked-reconstruction pre-training on current EEG corpora does not yet produce representations that transfer universally across BCI paradigms.

What carries the argument

The load-bearing mechanism is the benchmark itself: a standardized evaluation protocol applying leave-one-subject-out and within-subject few-shot scenarios, full-parameter fine-tuning and linear probing, across 13 public datasets spanning nine BCI paradigms. Aggregate ranking tables (Table VII and Fig. 8) and top-1/top-3 counts across tasks are the instruments that produce the central conclusion. The paper also contributes a taxonomy of pre-training objectives (masked raw-signal, token, frequency-domain, codebook, autoregressive) used to organize the 50 surveyed models.

Load-bearing premise

The scale conclusion assumes the cross-model comparison isolates model size, but the 12 foundation models differ simultaneously in pre-training data, architecture, objective, compute, and release date; the paper's own survey notes that capacity and resources vary non-monotonically (Section II-A).

What would settle it

Re-run the benchmark after removing all entries marked '*' (models pre-trained on the same downstream datasets) and check whether specialist models still outrank foundation models; alternatively, train the same architecture and pre-training objective with increasing data volumes and parameter counts under matched compute, and test whether downstream performance improves monotonically – if it does, the 'scale does not help' claim is overturned.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If foundation models require full-parameter fine-tuning anyway, their advantage over from-scratch specialists narrows to initialization, which current results suggest is often negative.
  • Benchmarking practice should adopt the two-scenario protocol (LOSO plus few-shot calibration) as a standard, since LOSO alone overstates deployment readiness.
  • Research priority should shift from scaling parameters and data toward pre-training objectives and data curation; the paper explicitly points to data quality as a likely limiting factor.
  • Paradigm-specific foundation models (e.g., MIRepNet for motor imagery) may offer a more practical route than universal models, because the target paradigm is usually known before deployment.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The 'larger is not better' conclusion is an observation about confounded comparisons, not a tested scaling law; a controlled study varying only model size while holding data, objective, and compute fixed would be needed to settle whether EEG exhibits scaling behavior.
  • The benchmark includes models pre-trained on the same downstream datasets (marked with '*' in the tables), which may inflate those models' scores; a re-ranking that excludes such entries would test how much of the specialist advantage is due to data overlap.
  • If the findings generalize, the BCI community may converge on compact, paradigm-tuned architectures rather than billion-parameter EEG transformers, lowering compute barriers for real-world deployment.
  • The few-shot analysis suggests relative model rankings are largely stable across calibration data volumes, implying that model selection could be performed with small calibration sets before committing to full fine-tuning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper presents EEG-FM-Compass, a survey and benchmark for EEG foundation models (FMs). It reviews roughly 50 recent models and organizes them into a taxonomy spanning preprocessing, architecture, and self-supervised objectives. It then evaluates 12 open-source FMs and 7 specialist baselines on 13 EEG datasets covering 9 BCI paradigms under two adaptation scenarios: leave-one-subject-out (LOSO) and within-subject few-shot. In each scenario it compares full-parameter fine-tuning against linear probing, examines fine-tuning data ratios, and analyzes model-scale effects. The three headline conclusions are: (1) linear probing is frequently insufficient relative to full fine-tuning; (2) specialist models trained from scratch remain highly competitive; and (3) larger FMs do not necessarily generalize better.

Significance. If the results are taken with appropriate caveats, this is a valuable resource for the EEG-FM community. The survey component is broad and the unified taxonomy is useful. The benchmark is unusually comprehensive in dataset coverage (13 datasets, 9 paradigms) and includes two deployment-oriented protocols, per-subject results, and mean/std over three seeds. The comparison of linear probing versus full fine-tuning is a useful empirical data point, and the finding that lightweight specialists such as EEGNet remain competitive is a meaningful caution for the field. The main weakness is that the scaling-law conclusion is over-interpreted from a confounded cross-model comparison, and the benchmark rankings are not adjusted for pre-training/downstream dataset overlap. These issues are fixable by reframing or additional analysis.

major comments (3)
  1. [Section III-D3, Table VII, Fig. 8, Section V] The central claim that 'increasing model size alone does not guarantee improved generalization' is presented as a tested scaling-law result, but the evidence is a cross-model rank comparison. The 12 FMs and 7 specialists differ simultaneously in parameter count, pre-training corpus (e.g., 1.5 TB vs 27,062 h vs 357,000 h), pre-training objective (masked raw reconstruction, codebook, contrastive, autoregressive), backbone (Transformer/Mamba/CNN), fine-tuning protocol, and release date. Section II-A itself notes that 'model capacity and training resources ... exhibit substantial variability rather than a monotonic scaling trend.' Table VII and Fig. 8 can support a descriptive observation about currently released checkpoints, not a causal claim about model size per se. This is load-bearing because Q3 and Section IV-C1 use the absence of scaling to motivate large-scale data construction. Plea
  2. [Section III-D, Tables V-VI] Several starred entries indicate that the pre-training corpus of a model overlaps with the downstream evaluation dataset or its parent dataset. For example, BENDR, CBraMod, EEGMamba, and LUNA are pre-trained on TUEG and evaluated on TUAB; BIOT-6D and TFM are pre-trained on CHB-MIT and evaluated on CHB-MIT; LaBraM and EEGPT are pre-trained on SEED and evaluated on SEED. The aggregate rankings in Table VII and Fig. 7 include these overlapping results without adjustment, so the 'transfer' score conflates generalization with near-distribution or in-distribution evaluation. This can bias the FM-versus-specialist comparison and the overall model rankings. Please either exclude overlapping (model, dataset) pairs from aggregate ranks, report them in a separate column, or conduct a sensitivity analysis restricted to non-overlapping tasks.
  3. [Table VII] The 'average rank' in Table VII is computed over models that are not evaluated on identical datasets: BrainOmni-Tiny/Base have '—' entries for CHB-MIT (Table V) and Sleep-EDFx (Table VI), and other missing entries exist. If ranks are averaged only over available datasets, models evaluated on a smaller or easier subset can obtain an advantage. The paper should state clearly whether each average is over the common set of tasks/scenarios or over the available entries per model. If the latter, the comparability of Table VII is compromised. Since this table is direct evidence for the top-level ranking and the scale conclusion, this issue should be corrected or the missing entries handled explicitly.
minor comments (5)
  1. [Abstract vs Introduction] The abstract says '55 representative models' while the introduction and Section II say '50 models'. Please reconcile the count.
  2. [Figure 3] The axis labels and legend entries in Figure 3 appear garbled (e.g., '/uni00000019/...' sequences). The figure is effectively unreadable in the submitted version and needs to be regenerated.
  3. [Table II] The 'Overlap' column uses symbols such as '%', '!', and '\%' without a legend. Please define these markers in the table caption or main text.
  4. [Section III-C, Appendix C] MIRepNet is listed among the evaluated foundation models in Section III-C, but it does not appear in the main Tables V-VI; it appears only in the appendix on MI datasets. This should be stated explicitly in the main text to avoid confusion about the 12-model benchmark.
  5. [References] Reference [33] is titled 'EEGPT: Unleashing the potential of EEG generalist foundation model by autoregressive pre-training' but the corresponding row in Table I is labeled 'BrainGPT'. Please correct the citation/title mismatch.

Circularity Check

0 steps flagged

No significant circularity: benchmark conclusions are empirical; self-citations are minor and not load-bearing.

full rationale

EEG-FM-Compass is a benchmark and survey rather than a derivation, so the main circularity patterns do not apply. The headline results — linear probing is frequently insufficient, specialist models remain competitive, and larger FMs do not necessarily generalize better — are obtained by fine-tuning released checkpoints and baselines on fixed LOSO and within-subject few-shot protocols, then reading accuracies and ranks from Tables V–VII and Figs. 7–8. No parameter is fitted to the benchmark and then reported as a prediction; the rankings are direct empirical measurements. The scaling claim is admittedly descriptive: Section II-A itself states that 'model capacity and training resources ... exhibit substantial variability rather than a monotonic scaling trend,' and the cross-model comparison is confounded by differing pre-training data, objectives, backbones, compute, and release dates. That confound is a validity caveat, not a circular reduction, because the paper does not claim to have isolated model size by construction. The paper does contain minor self-citations: MIRepNet [49] is the authors' own paradigm-specific model and is highlighted in Section IV-A and the appendix, while CLEAN-MI [71] and EA references [72]–[73] involve the same group. However, MIRepNet is not in the main Table VII overall ranking or in the specialist-vs-FM top-1/top-3 aggregates that drive the central conclusions; it appears only in the MI appendix and discussion. CLEAN-MI is cited only to motivate future data-quality research. Thus the self-citations are not load-bearing, and there is no equation-level or construction-level circularity. Score 2 reflects these minor, non-load-bearing self-references rather than any substantive circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

The paper is an empirical benchmark, so it introduces no free parameters in the sense of fitted theory. Its load-bearing assumptions are about protocol fairness, checkpoint comparability, and the treatment of pre-training overlap with downstream datasets. The paper does not invent new physical or architectural entities; MIRepNet is a previously proposed model from the same group.

axioms (3)
  • domain assumption The benchmark protocols (LOSO and within-subject few-shot) are representative of real-world BCI deployment and are fair to both foundation and specialist models.
    Section III-B defines the two scenarios; if these protocols systematically disadvantage foundation models (e.g., through insufficient fine-tuning budgets or unfavorable hyperparameters), the main conclusions would not generalize to deployment.
  • domain assumption Publicly released checkpoints and default fine-tuning procedures provide a fair comparison of the model families.
    Section III-C evaluates released models without evidence of per-model hyperparameter search; differences in fine-tuning difficulty could be mistaken for differences in representation quality.
  • domain assumption Including evaluation datasets that overlap with pre-training (marked '*') does not materially distort the aggregate rankings.
    Tables V-VI mark datasets used in pre-training but do not exclude them from ranking or adjust for this home-field advantage, which could inflate the performance of some foundation models.

pith-pipeline@v1.3.0-alltime-deepseek · 77379 in / 7592 out tokens · 96091 ms · 2026-08-04T06:14:56.231749+00:00 · methodology

0 comments
read the original abstract

Electroencephalography (EEG) foundation models (FMs) have recently emerged as a promising paradigm for brain-computer interfaces, aiming to learn transferable neural representations from large-scale heterogeneous recordings. Despite rapid progress, a fair and comprehensive comparison of existing EEG FMs is still lacking, owing to inconsistent pre-training objectives, preprocessing choices, and downstream evaluation protocols. To fill this gap, we present EEG-FM-Compass. We first review 55 representative models and organize their design choices into a unified taxonomic framework including data standardization, model architectures, and self-supervised pre-training strategies. We then evaluate 12 open source FMs and competitive specialist baselines across 13 EEG datasets spanning nine brain-computer interface paradigms. Emphasizing real-world deployments, we consider both cross-subject generalization under a leave-one-subject-out protocol and rapid calibration under a within-subject few-shot setting. We further compare full-parameter fine-tuning with linear probing to assess the transferability of pre-trained representations, and examine the relationship between model scale and downstream performance. Our results indicate that: 1) linear probing is frequently insufficient; 2) specialist models trained from scratch remain competitive across many tasks; and 3) larger FMs do not necessarily yield better generalization performance under current data regimes and training practices.

Figures

Figures reproduced from arXiv: 2601.17883 by Dingkun Liu, Dongrui Wu, Jiayu An, Jingwei Luo, Yaozhi Wen, Yuheng Chen, Zhenyao Cui, Zhu Chen.

Figure 1
Figure 1. Figure 1: Overview of BCI foundation models. Models are pre [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: EEG foundation model pre-training pipeline. Raw EEG trials are first standardized through channel selection or [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of 50 existing EEG foundation models. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Dataset usage statistics across existing EEG foundation models. (a) Frequency ranking of datasets used during pre [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Overview of datasets and evaluation scenarios used in the benchmark. (a) The 13 downstream datasets spanning 9 [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: t-SNE visualization of the BNCI2014008 and SEED datasets. achieved the highest overall ranking with an average rank of 5.96 and a model size of 4.0M parameters. EEGNet, a widely utilized lightweight EEG decoding backbone, attained second place with only 2K parameters. These results demonstrated that larger models do not necessarily yield better performance. This observation may be attributed to two factors… view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of ranking performance between specialist models and foundation models. (a) and (b) show the number [PITH_FULL_IMAGE:figures/full_fig_p019_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Overall ranking of EEG foundation models with respect to release date and model size (bubble size indicates parameter [PITH_FULL_IMAGE:figures/full_fig_p020_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Performance comparison across different fine-tuning data ratios on the BNCI2014001 dataset (MI paradigm) and the [PITH_FULL_IMAGE:figures/full_fig_p021_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: t-SNE visualization of EEG trials from the BNCI2014004 dataset. (a) Before EA; (b) After EA. Different colors represent trials from different subjects. that facilitate rapid adaptation to new tasks, including methods to achieve competitive performance with less calibration data and techniques to accelerate distribution alignment between pre-trained models and downstream data. V. CONCLUSION This paper has … view at source ↗
Figure 11
Figure 11. Figure 11: Impact of EA on model performance on the BNCI2014001 dataset. [PITH_FULL_IMAGE:figures/full_fig_p023_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Accuracy comparison on the BNCI2014001 dataset across different fine-tuning ratios. [PITH_FULL_IMAGE:figures/full_fig_p031_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Accuracy comparison on the BNCI2014004 dataset across different fine-tuning ratios. [PITH_FULL_IMAGE:figures/full_fig_p066_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Accuracy comparison on the BNCI2015001 dataset across different fine-tuning ratios. [PITH_FULL_IMAGE:figures/full_fig_p067_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Balanced classification accuracy comparison on the BNCI2014009 dataset across different fine-tuning ratios. [PITH_FULL_IMAGE:figures/full_fig_p068_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Balanced classification accuracy comparison on the BNCI2014008 dataset across different fine-tuning ratios. [PITH_FULL_IMAGE:figures/full_fig_p069_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Accuracy comparison on the Nakanishi2015 dataset across different fine-tuning ratios. [PITH_FULL_IMAGE:figures/full_fig_p070_17.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 7 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. EEG-FM-Audit: A Systematic Evaluation and Analysis Pipeline for EEG Foundation Models

    cs.LG 2026-05 unverdicted novelty 7.0

    EEG-FM-Audit is an evaluation pipeline showing that properly tuned supervised baselines can match or outperform EEG foundation models with far fewer parameters on public datasets.

  2. Aperiodic and Low-Frequency Spectral Bias in Reconstruction based EEG Foundation Models

    cs.LG 2026-05 unverdicted novelty 7.0

    Reconstruction-based EEG foundation models preferentially encode aperiodic and low-frequency components over oscillatory structure, with embeddings capturing subject identity more than task-relevant information.

  3. NeuroAtlas: Benchmarking Foundation Models for Clinical EEG and Brain-Computer Interfaces

    cs.LG 2026-05 unverdicted novelty 7.0

    NeuroAtlas benchmarks foundation models on 42 EEG datasets and reports that EEG-specific models do not consistently outperform generic time-series models, standard metrics miss clinical utility, and rankings vary by domain.

  4. NeuralBench: A Unifying Framework to Benchmark NeuroAI Models

    cs.LG 2026-05 conditional novelty 7.0

    NeuralBench is a new benchmarking framework for neuroAI models on EEG data that finds foundation models only marginally outperform task-specific ones while many tasks like cognitive decoding stay highly challenging.

  5. Foundation Model Guided Dual-Branch Co-Adaptation for Source-Free EEG Decoding

    eess.SP 2026-04 unverdicted novelty 7.0

    FUSED integrates EEG foundation models into source-free domain adaptation via dual-branch co-adaptation, consensus filtering, and two-stage pseudo-label refinement to achieve state-of-the-art cross-subject EEG decoding.

  6. Dive into Waves: Morlet Spectral Transformer for Cross-Subject Emotion Decoding from EEG

    cs.LG 2026-05 unverdicted novelty 6.0

    MST combines Morlet wavelet tokenization, long-context baseline removal, and frequency-specific spatial projections with a Transformer backbone to outperform pretrained EEG models on SEED datasets for cross-subject em...

  7. From Clever Hans to Scientific Discovery: Interpreting EEG Foundational Transformers with LRP

    cs.AI 2026-05 unverdicted novelty 6.0

    LRP on EEG transformers reveals Clever Hans artifacts in motor imagery tasks and a recurring central electrode cluster as a candidate sensorimotor signature of arousal.

Reference graph

Works this paper leans on

86 extracted references · 16 linked inside Pith · cited by 7 Pith papers

  1. [1]

    Brain computer interfaces, a review,

    L. F. Nicolas-Alonso and J. Gomez-Gil, “Brain computer interfaces, a review,”Sensors, vol. 12, no. 2, pp. 1211–1279, 2012

    2012

  2. [2]

    Canine EEG helps human: Cross- species and cross-modality epileptic seizure detection via multi-space alignment,

    Z. Wang, S. Li, and D. Wu, “Canine EEG helps human: Cross- species and cross-modality epileptic seizure detection via multi-space alignment,”National Science Review, vol. 12, no. 6, p. nwaf086, 2025

    2025

  3. [3]

    Multimodal BCIs: Target detection, multidimensional control, and awareness evalu- ation in patients with disorder of consciousness,

    Y . Li, J. Pan, J. Long, T. Yu, F. Wang, Z. Yu, and W. Wu, “Multimodal BCIs: Target detection, multidimensional control, and awareness evalu- ation in patients with disorder of consciousness,”Proc. IEEE, vol. 104, no. 2, pp. 332–352, 2016

    2016

  4. [4]

    Affective brain-computer interfaces (aBCIs): A tutorial,

    D. Wu, B.-L. Lu, B. Hu, and Z. Zeng, “Affective brain-computer interfaces (aBCIs): A tutorial,”Proc. of the IEEE, vol. 11, no. 10, pp. 1314–1332, 2023

    2023

  5. [5]

    Artificial intelli- gence as an emerging technology in the current care of neurological disorders,

    U. K. Patel, A. Anwar, S. Saleem, P. Malik, B. Rasul, K. Patel, R. Yao, A. Seshadri, M. Yousufuddin, and K. Arumaithurai, “Artificial intelli- gence as an emerging technology in the current care of neurological disorders,”Journal of Neurology, vol. 268, no. 5, pp. 1623–1642, 2021

    2021

  6. [6]

    Comparative analysis to identify efficient technique for interfacing BCI system,

    M. K. Kumar, B. Parameshachari, S. Prabu, and S. liberata Ullo, “Comparative analysis to identify efficient technique for interfacing BCI system,” inIOP Conference Series: Materials Science and Engineering, vol. 925, no. 1. IOP Publishing, 2020, p. 012062

    2020

  7. [7]

    A neuroergonomics approach to mental workload, engagement and human performance,

    F. Dehais, A. Lafont, R. Roy, and S. Fairclough, “A neuroergonomics approach to mental workload, engagement and human performance,” Frontiers in Neuroscience, vol. 14, p. 268, 2020

    2020

  8. [8]

    Imagined speech can be decoded from low-and cross-frequency intracranial EEG features,

    T. Proix, J. Delgado Saa, A. Christen, S. Martin, B. N. Pasley, R. T. Knight, X. Tian, D. Poeppel, W. K. Doyle, O. Devinskyet al., “Imagined speech can be decoded from low-and cross-frequency intracranial EEG features,”Nature Communications, vol. 13, no. 1, p. 48, 2022

    2022

  9. [9]

    Magnetoen- cephalography (MEG) based non-invasive Chinese speech decoding,

    Z. Jia, H. Wang, Y . Shen, F. Hu, J. An, K. Shu, and D. Wu, “Magnetoen- cephalography (MEG) based non-invasive Chinese speech decoding,” Journal of Neural Engineering, vol. 22, p. 066014, 2025

    2025

  10. [10]

    EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces,

    V . J. Lawhern, A. J. Solon, N. R. Waytowich, S. M. Gordon, C. P. Hung, and B. J. Lance, “EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces,”Journal of Neural Engineering, vol. 15, no. 5, p. 056013, 2018

    2018

  11. [11]

    EEG-based subject-independent emotion recognition using gated recurrent unit and minimum class confusion,

    H. Cui, A. Liu, X. Zhang, X. Chen, J. Liu, and X. Chen, “EEG-based subject-independent emotion recognition using gated recurrent unit and minimum class confusion,”IEEE Trans. on Affective Computing, vol. 14, no. 4, pp. 2740–2750, 2023

    2023

  12. [12]

    DBConformer: Dual-branch convolutional transformer for EEG decoding,

    Z. Wang, H. Wang, T. Jia, X. He, S. Li, and D. Wu, “DBConformer: Dual-branch convolutional transformer for EEG decoding,”IEEE Jour- nal of Biomedical and Health Informatics, 2026, in press

    2026

  13. [13]

    SDDA: Spatial distillation based distribution alignment for cross-headset EEG classification,

    D. Liu, S. Li, Z. Wang, W. Li, and D. Wu, “SDDA: Spatial distillation based distribution alignment for cross-headset EEG classification,”IEEE Trans. on Biomedical Engineering, 2025

    2025

  14. [14]

    AFPM: Alignment-based frame patch modeling for cross-dataset EEG decoding,

    X. Chen, S. Li, and D. Wu, “AFPM: Alignment-based frame patch modeling for cross-dataset EEG decoding,”Science China Information Sciences, 2026, in press

    2026

  15. [15]

    A comprehensive overview of large language models,

    H. Naveed, A. U. Khan, S. Qiu, M. Saqib, S. Anwar, M. Usman, N. Akhtar, N. Barnes, and A. Mian, “A comprehensive overview of large language models,”ACM Trans. on Intelligent Systems and Technology, vol. 16, no. 5, pp. 1–72, 2025. 23 /uni00000028/uni00000028/uni0000002a/uni00000031/uni00000048/uni00000057 /uni00000036/uni0000004b/uni00000044/uni0000004f/un...

    2025

  16. [16]

    Toward the unification of generative and discriminative visual foundation model: a survey,

    X. Liu, T. Zhou, C. Wang, Y . Wang, Y . Wang, Q. Cao, W. Du, Y . Yang, J. He, Y . Qiaoet al., “Toward the unification of generative and discriminative visual foundation model: a survey,”The Visual Computer, vol. 41, no. 5, pp. 3371–3412, 2025

    2025

  17. [17]

    Foundation models for EEG decoding: current progress and prospective research,

    Y . Yuxuan, W. Hongbo, C. Li, P. Yiheng, and J. Luo, “Foundation models for EEG decoding: current progress and prospective research,”Journal of Neural Engineering, 2025

    2025

  18. [18]

    BENDR: Using trans- formers and a contrastive self-supervised learning task to learn from massive amounts of EEG data,

    D. Kostas, S. Aroca-Ouellette, and F. Rudzicz, “BENDR: Using trans- formers and a contrastive self-supervised learning task to learn from massive amounts of EEG data,”Frontiers in Human Neuroscience, vol. 15, p. 653659, 2021

    2021

  19. [19]

    BrainBERT: Self-supervised representation learning for intracranial recordings,

    C. Wang, V . Subramaniam, A. U. Yaari, G. Kreiman, B. Katz, I. Cases, and A. Barbu, “BrainBERT: Self-supervised representation learning for intracranial recordings,” Kigali, Rwanda, May 2023

    2023

  20. [20]

    Mbrain: A multi-channel self-supervised learning framework for brain signals,

    D. Cai, J. Chen, Y . Yang, T. Liu, and Y . Li, “Mbrain: A multi-channel self-supervised learning framework for brain signals,” inProc. of the 29th ACM SIGKDD Conf. on Knowledge Discovery and Data Mining, Long Beach, CA, Aug. 2023, pp. 130–141

    2023

  21. [21]

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

    C. Yang, M. Westover, and J. Sun, “BIOT: Biosignal transformer for cross-data learning in the wild,”Advances in Neural Information Processing Systems, vol. 36, pp. 78 240–78 260, Dec. 2023

    2023

  22. [22]

    Brant: Foundation model for intracranial neural signal,

    D. Zhang, Z. Yuan, Y . Yang, J. Chen, J. Wang, and Y . Li, “Brant: Foundation model for intracranial neural signal,”Advances in Neural Information Processing Systems, vol. 36, pp. 26 304–26 321, Dec. 2023

    2023

  23. [23]

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

    W. Jiang, L. Zhao, and B.-l. Lu, “Large brain model for learning generic representations with tremendous EEG data in BCI,” inThe Twelfth Int’l Conf. on Learning Representations, Vienna, Austria, May 2024

    2024

  24. [24]

    Mentality: A mamba- based approach towards foundation models for EEG,

    S. Panchavati, C. Arnold, and W. Speier, “Mentality: A mamba- based approach towards foundation models for EEG,”arXiv preprint arXiv:2509.02746, 2025

    Pith/arXiv arXiv 2025

  25. [25]

    Neuro-GPT: Towards a foundation model for EEG,

    W. Cui, W. Jeong, P. Th ¨olke, T. Medani, K. Jerbi, A. A. Joshi, and R. M. Leahy, “Neuro-GPT: Towards a foundation model for EEG,” inIEEE Int’l Symposium on Biomedical Imaging (ISBI). IEEE, 2024, pp. 1–5

    2024

  26. [26]

    MEET: A multi-band EEG transformer for brain states decoding,

    E. Shi, S. Yu, Y . Kang, J. Wu, L. Zhao, D. Zhu, J. Lv, T. Liu, X. Hu, and S. Zhang, “MEET: A multi-band EEG transformer for brain states decoding,”IEEE Trans. on Biomedical Engineering, vol. 71, no. 5, pp. 1442–1453, 2023

    2023

  27. [27]

    EEGFormer: Towards transferable and interpretable large-scale EEG foundation model,

    Y . Chen, K. Ren, K. Song, Y . Wang, Y . Wang, D. Li, and L. Qiu, “EEGFormer: Towards transferable and interpretable large-scale EEG foundation model,” inAAAI 2024 Spring Symposium on Clinical Foun- dation Models, Stanford, CA, Mar. 2024

    2024

  28. [28]

    Brainwave: A brain signal foundation model for clinical applications,

    Z. Yuan, F. Shen, M. Li, Y . Yu, C. Tan, and Y . Yang, “Brainwave: A brain signal foundation model for clinical applications,”arXiv preprint arXiv:2402.10251, 2024

    Pith/arXiv arXiv 2024

  29. [29]

    NeuroLM: A universal multi- task foundation model for bridging the gap between language and EEG signals,

    W. Jiang, Y . Wang, B.-l. Lu, and D. Li, “NeuroLM: A universal multi- task foundation model for bridging the gap between language and EEG signals,” inThe Thirteenth Int’l Conf. on Learning Representations, Vienna, Austria, May. 2024

    2024

  30. [30]

    Brant-X: A unified physiological signal alignment framework,

    D. Zhang, Z. Yuan, J. Chen, K. Chen, and Y . Yang, “Brant-X: A unified physiological signal alignment framework,” inProc. of the 30th ACM SIGKDD Conf. on Knowledge Discovery and Data Mining, Barcelona, Spain, Aug. 2024, pp. 4155–4166

    2024

  31. [31]

    FoME: A foundation model for EEG using adaptive temporal-lateral attention scaling,

    E. Shi, K. Zhao, Q. Yuan, J. Wang, H. Hu, S. Yu, and S. Zhang, “FoME: A foundation model for EEG using adaptive temporal-lateral attention scaling,”arXiv preprint arXiv:2409.12454, 2024

    Pith/arXiv arXiv 2024

  32. [32]

    EEGPT: Pre- trained transformer for universal and reliable representation of EEG signals,

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

    2024

  33. [33]

    EEGPT: Unleashing the potential of EEG generalist foundation model by autoregressive pre-training,

    T. Yue, S. Xue, X. Gao, Y . Tang, L. Guo, J. Jiang, and J. Liu, “EEGPT: Unleashing the potential of EEG generalist foundation model by autoregressive pre-training,”arXiv preprint arXiv:2410.19779, 2024

    Pith/arXiv arXiv 2024

  34. [34]

    GEFM: Graph-enhanced EEG foundation model,

    L. Wang, T. Suzumura, and H. Kanezashi, “GEFM: Graph-enhanced EEG foundation model,” in2025 47th Annual Int’l Conf. of the IEEE Engineering in Medicine and Biology Society (EMBC). IEEE, 2025, pp. 1–7

    2025

  35. [35]

    CBramod: A criss-cross brain foundation model for EEG decoding,

    J. Wang, S. Zhao, Z. Luo, Y . Zhou, H. Jiang, S. Li, T. Li, and G. Pan, “CBramod: A criss-cross brain foundation model for EEG decoding,” inThe Thirteenth Int’l Conf. on Learning Representations, Singapore, Apr. 2025

    2025

  36. [36]

    CERebro: Compact encoder for representations of brain oscillations using efficient alternating attention,

    A. Dimofte, G. A. Bucagu, T. M. Ingolfsson, X. Wang, A. Cossettini, L. Benini, and Y . Li, “CERebro: Compact encoder for representations of brain oscillations using efficient alternating attention,”arXiv preprint arXiv:2501.10885, 2025

    arXiv 2025

  37. [37]

    LEAD: Large foundation model for EEG-based alzheimer’s disease detection,

    Y . Wang, N. Huang, N. Mammone, M. Cecchi, and X. Zhang, “LEAD: Large foundation model for EEG-based alzheimer’s disease detection,” arXiv preprint arXiv:2502.01678, 2025

    arXiv 2025

  38. [38]

    FEMBA: Efficient and scalable EEG analysis with a bidirectional mamba foun- dation model,

    A. Tegon, T. M. Ingolfsson, X. Wang, L. Benini, and Y . Li, “FEMBA: Efficient and scalable EEG analysis with a bidirectional mamba foun- dation model,”arXiv preprint arXiv:2502.06438, 2025

    arXiv 2025

  39. [39]

    Large cognition model: Towards pretrained EEG foundation model,

    C.-S. Chen, Y .-J. Chen, and A. H.-W. Tsai, “Large cognition model: Towards pretrained EEG foundation model,”arXiv preprint arXiv:2502.17464, 2025

    Pith/arXiv arXiv 2025

  40. [40]

    Tokenizing single- channel EEG with time-frequency motif learning,

    J. Pradeepkumar, X. Piao, Z. Chen, and J. Sun, “Tokenizing single- channel EEG with time-frequency motif learning,” inNeurIPS 2025 Workshop on Learning from Time Series for Health, San Diego, CA, Dec. 2025

    2025

  41. [41]

    ALFEE: Adap- tive large foundation model for EEG representation,

    W. Xiong, J. Lin, J. Li, J. Li, and C. Jiang, “ALFEE: Adap- tive large foundation model for EEG representation,”arXiv preprint arXiv:2505.06291, 2025

    Pith/arXiv arXiv 2025

  42. [42]

    Brainomni: A brain foundation model 24 for unified EEG and MEG signals,

    Q. Xiao, Z. Cui, C. Zhang, S. Chen, W. Wu, A. Thwaites, A. Wool- gar, B. Zhou, and C. Zhang, “Brainomni: A brain foundation model 24 for unified EEG and MEG signals,”Advances in Neural Information Processing Systems, Dec. 2025

    2025

  43. [43]

    EEG foundation models for BCI learn diverse features of electrophysiology,

    M. Ogg, R. Hingorani, D. Luna, G. W. Milsap, W. G. Coon, and C. A. Scholl, “EEG foundation models for BCI learn diverse features of electrophysiology,”arXiv preprint arXiv:2506.01867, 2025

    Pith/arXiv arXiv 2025

  44. [44]

    Codebrain: Towards decoupled interpretability and multi-scale architecture for EEG foundation model,

    J. Ma, F. Wu, Q. Lin, Y . Xing, C. Liu, Z. Jia, and M. Feng, “Codebrain: Towards decoupled interpretability and multi-scale architecture for EEG foundation model,”arXiv preprint arXiv:2506.09110, 2025

    Pith/arXiv arXiv 2025

  45. [45]

    Unimind: Unleashing the power of LLMs for unified multi- task brain decoding,

    W. Lu, C. Song, J. Wu, P. Zhu, Y . Zhou, W. Mai, Q. Zheng, and W. Ouyang, “Unimind: Unleashing the power of LLMs for unified multi- task brain decoding,”arXiv preprint arXiv:2506.18962, 2025

    Pith/arXiv arXiv 2025

  46. [46]

    CSBrain: A cross-scale spatiotemporal brain foundation model for EEG decoding,

    Y . Zhou, J. Wu, Z. Ren, Z. Yao, W. Lu, K. Peng, Q. Zheng, C. Song, W. Ouyang, and C. Gou, “CSBrain: A cross-scale spatiotemporal brain foundation model for EEG decoding,”Advances in Neural Information Processing Systems, Dec. 2025

    2025

  47. [47]

    DMAE-EEG: A pretraining framework for EEG spatiotemporal representation learning,

    Y . Zhang, Y . Yu, H. Li, A. Wu, X. Chen, J. Liu, L.-L. Zeng, and D. Hu, “DMAE-EEG: A pretraining framework for EEG spatiotemporal representation learning,”IEEE Trans. on Neural Networks and Learning Systems, 2025

    2025

  48. [48]

    EEGMamba: An EEG foundation model with mamba,

    J. Wang, S. Zhao, Z. Luo, Y . Zhou, S. Li, and G. Pan, “EEGMamba: An EEG foundation model with mamba,”Neural Networks, p. 107816, 2025

    2025

  49. [49]

    MIRepnet: A pipeline and foundation model for EEG-based motor imagery classification,

    D. Liu, Z. Chen, J. Luo, S. Lian, and D. Wu, “MIRepnet: A pipeline and foundation model for EEG-based motor imagery classification,”arXiv preprint arXiv:2507.20254, 2025

    Pith/arXiv arXiv 2025

  50. [50]

    Foundation models reveal untapped health information in human polysomnographic sleep data,

    W. G. Coon and M. Ogg, “Foundation models reveal untapped health information in human polysomnographic sleep data,”medRxiv, pp. 2025–07, 2025

    2025

  51. [51]

    EEGDM: EEG representation learning via generative diffusion model,

    J. H. Puah, S. K. Goh, Z. Zhang, Z. Ye, C. K. Chan, K. S. Lim, S. L. Fong, K. S. Woon, and C. Guan, “EEGDM: EEG representation learning via generative diffusion model,”arXiv preprint arXiv:2508.14086, 2025

    Pith/arXiv arXiv 2025

  52. [52]

    CoMET: A contrastive-masked brain foundation model for universal EEG representation,

    A. Li, Z. Wang, L. Yang, Z. Wang, T. Xu, H. Hu, and M. M. Van Hulle, “CoMET: A contrastive-masked brain foundation model for universal EEG representation,”arXiv preprint arXiv:2509.00314, 2025

    Pith/arXiv arXiv 2025

  53. [53]

    EpilepsyFM: A domain-specific foundation model for epileptic representation learning using EEG signals,

    Z. Li, N. Zhu, Y . Chen, B. Chen, Q. Dong, L. Gan, S. Zhao, Z. Yan, and T. Zhang, “EpilepsyFM: A domain-specific foundation model for epileptic representation learning using EEG signals,”Neural Networks, p. 108060, 2025

    2025

  54. [54]

    SingLEM: Single-channel large EEG model,

    J. Sukhbaatar, S. Imamura, I. Inoue, S. Murakami, K. M. Hassan, S. Han, I. Chanpornpakdi, and T. Tanaka, “SingLEM: Single-channel large EEG model,”arXiv preprint arXiv:2509.17920, 2025

    Pith/arXiv arXiv 2025

  55. [55]

    Brainpro: Towards large-scale brain state-aware EEG representation learning,

    Y . Ding, M. Jiang, W. Jiang, S. Zhang, X. Zhou, C. Liu, S. Li, Y . Li, and C. Guan, “Brainpro: Towards large-scale brain state-aware EEG representation learning,”arXiv preprint arXiv:2509.22050, 2025

    Pith/arXiv arXiv 2025

  56. [56]

    Uni-NTFM: A unified foundation model for eeg signal representation learning,

    Z. Chen, Y . Zhang, Q. Lan, T. Liu, H. Wang, Y . Ding, Z. Jia, R. Chen, K. Wang, and X. Zhou, “Uni-NTFM: A unified foundation model for eeg signal representation learning,”arXiv preprint arXiv:2509.24222, 2025

    arXiv 2025

  57. [57]

    ELASTIQ: EEG-language align- ment with semantic task instruction and querying,

    M. Jiang, S. Zhang, Z. Yang, M. Wu, W. Jiang, Z. Guo, W. Zhang, R. Liu, S. Zhang, Y . Liet al., “ELASTIQ: EEG-language align- ment with semantic task instruction and querying,”arXiv preprint arXiv:2509.24302, 2025

    arXiv 2025

  58. [58]

    Neural codecs as biosignal tokenizers,

    K. Avramidis, T. Feng, W. Jeong, J. Lee, W. Cui, R. M. Leahy, and S. Narayanan, “Neural codecs as biosignal tokenizers,”arXiv preprint arXiv:2510.09095, 2025

    arXiv 2025

  59. [59]

    HEAR: An EEG foundation model with heterogeneous electrode adaptive representation,

    Z. Chen, C. Qin, W. You, R. Liu, C. Chu, R. Yang, K. C. Tan, and J. Wu, “HEAR: An EEG foundation model with heterogeneous electrode adaptive representation,”arXiv preprint arXiv:2510.12515, 2025

    arXiv 2025

  60. [60]

    NeuroRVQ: Multi-scale EEG to- kenization for generative large brainwave models,

    K. Barmpas, N. Lee, A. Koliousis, Y . Panagakis, D. A. Adamos, N. Laskaris, and S. Zafeiriou, “NeuroRVQ: Multi-scale EEG to- kenization for generative large brainwave models,”arXiv preprint arXiv:2510.13068, 2025

    Pith/arXiv arXiv 2025

  61. [61]

    REVE: A foundation model for EEG-adapting to any setup with large-scale pretraining on 25,000 subjects,

    Y . El Ouahidi, J. Lys, P. Th ¨olke, N. Farrugia, B. Pasdeloup, V . Gripon, K. Jerbi, and G. Lioi, “REVE: A foundation model for EEG-adapting to any setup with large-scale pretraining on 25,000 subjects,” inThe Thirty-ninth Annual Conf. on Neural Information Processing Systems, San Diego, CA, Dec. 2025

    2025

  62. [62]

    Multi-dataset joint pre- training of emotional EEG enables generalizable affective computing,

    Q. Zhang, J. Zhong, Z. Li, X. Shen, and Q. Liu, “Multi-dataset joint pre- training of emotional EEG enables generalizable affective computing,” arXiv preprint arXiv:2510.22197, 2025

    arXiv 2025

  63. [63]

    LUNA: Efficient and topology-agnostic foundation model for EEG signal analysis,

    B. D ¨oner, T. M. Ingolfsson, L. Benini, and Y . Li, “LUNA: Efficient and topology-agnostic foundation model for EEG signal analysis,”arXiv preprint arXiv:2510.22257, 2025

    arXiv 2025

  64. [64]

    THD-BAR: Topology hierarchical derived brain autoregressive modeling for EEG generic rep- resentations,

    W. Yang, W. Yan, W. Liu, Y . Ma, and Y . Li, “THD-BAR: Topology hierarchical derived brain autoregressive modeling for EEG generic rep- resentations,” inThe Thirty-ninth Annual Conf. on Neural Information Processing Systems, San Diego, CA, Dec. 2025

    2025

  65. [65]

    EEG-X: Device-agnostic and noise-robust foundation model for EEG,

    N. M. Foumani, S. Ghane, N. Nguyen, M. Salehi, G. I. Webb, and G. Mackellar, “EEG-X: Device-agnostic and noise-robust foundation model for EEG,”arXiv preprint arXiv:2511.08861, 2025

    arXiv 2025

  66. [66]

    SAMBA: Toward a long-context EEG foundation model via spatial embedding and differential mamba,

    J. Hong, G. Mackellar, and S. Ghane, “SAMBA: Toward a long-context EEG foundation model via spatial embedding and differential mamba,” arXiv preprint arXiv:2511.18571, 2025

    arXiv 2025

  67. [67]

    DeeperBrain: A neuro-grounded EEG foundation model towards universal BCI,

    J. Wang, S. Zhao, Y . Zhou, Y . Kang, S. Li, and G. Pan, “DeeperBrain: A neuro-grounded EEG foundation model towards universal BCI,”arXiv preprint arXiv:2601.06134, 2026

    arXiv 2026

  68. [68]

    Spatial and temporal resolutions of EEG: Is it really black and white? a scalp current density view,

    B. Burle, L. Spieser, C. Roger, L. Casini, T. Hasbroucq, and F. Vidal, “Spatial and temporal resolutions of EEG: Is it really black and white? a scalp current density view,”International Journal of Psychophysiology, vol. 97, no. 3, pp. 210–220, 2015

    2015

  69. [69]

    Foundation models: A new paradigm for artificial intelligence,

    J. Schneider, C. Meske, and P. Kuss, “Foundation models: A new paradigm for artificial intelligence,”Business & Information Systems Engineering, vol. 66, no. 2, pp. 221–231, 2024

    2024

  70. [70]

    Foundation models defining a new era in vision: a survey and outlook,

    M. Awais, M. Naseer, S. Khan, R. M. Anwer, H. Cholakkal, M. Shah, M.-H. Yang, and F. S. Khan, “Foundation models defining a new era in vision: a survey and outlook,”IEEE Trans. on Pattern Analysis and Machine Intelligence, 2025

    2025

  71. [71]

    CLEAN-MI: A scalable and efficient pipeline for constructing high-quality neurodata in motor imagery paradigm,

    D. Liu, Z. Chen, and D. Wu, “CLEAN-MI: A scalable and efficient pipeline for constructing high-quality neurodata in motor imagery paradigm,”arXiv preprint arXiv:2506.11830, 2025

    Pith/arXiv arXiv 2025

  72. [72]

    Transfer learning for brain-computer interfaces: A Euclidean space data alignment approach,

    H. He and D. Wu, “Transfer learning for brain-computer interfaces: A Euclidean space data alignment approach,”IEEE Trans. on Biomedical Engineering, vol. 67, no. 2, pp. 399–410, 2020

    2020

  73. [73]

    Revisiting Euclidean alignment for transfer learning in EEG-based brain-computer interfaces,

    D. Wu, “Revisiting Euclidean alignment for transfer learning in EEG-based brain-computer interfaces,”Journal of Neural Engineering, vol. 22, p. 031005, 2025

    2025

  74. [74]

    Optimizing spatial filters for robust EEG single-trial analysis,

    B. Blankertz, R. Tomioka, S. Lemm, M. Kawanabe, and K.-R. Muller, “Optimizing spatial filters for robust EEG single-trial analysis,”IEEE Signal Processing Magazine, vol. 25, no. 1, pp. 41–56, 2007

    2007

  75. [75]

    xDAWN algorithm to enhance evoked potentials: application to brain–computer interface,

    B. Rivet, A. Souloumiac, V . Attina, and G. Gibert, “xDAWN algorithm to enhance evoked potentials: application to brain–computer interface,” IEEE Trans. on Biomedical Engineering, vol. 56, no. 8, pp. 2035–2043, 2009

    2035

  76. [76]

    Classification of EEG evoked in 2d and 3d virtual reality: traditional machine learning versus deep learning,

    M. Zuo, B. Yu, and L. Sui, “Classification of EEG evoked in 2d and 3d virtual reality: traditional machine learning versus deep learning,” Biomedical Physics & Engineering Express, vol. 11, no. 1, p. 015005, 2024

    2024

  77. [77]

    Temporal feature extraction and machine learning for classification of sleep stages using telemetry polysomnography,

    U. Lal, S. Mathavu Vasanthsena, and A. Hoblidar, “Temporal feature extraction and machine learning for classification of sleep stages using telemetry polysomnography,”Brain Sciences, vol. 13, no. 8, p. 1201, 2023

    2023

  78. [78]

    Enhancing detection of SSVEPs for a high-speed brain speller using task-related component analysis,

    M. Nakanishi, Y . Wang, X. Chen, Y .-T. Wang, X. Gao, and T.-P. Jung, “Enhancing detection of SSVEPs for a high-speed brain speller using task-related component analysis,”IEEE Trans. on Biomedical Engineering, vol. 65, no. 1, pp. 104–112, 2017

    2017

  79. [79]

    Multimodal vigilance estimation using deep learning,

    W. Wu, W. Sun, Q. J. Wu, Y . Yang, H. Zhang, W.-L. Zheng, and B.-L. Lu, “Multimodal vigilance estimation using deep learning,”IEEE Trans. on Cybernetics, vol. 52, no. 5, pp. 3097–3110, 2020

    2020

  80. [80]

    Deep learning with convolutional neural networks for EEG decoding and visualization,

    R. T. Schirrmeister, J. T. Springenberg, L. D. J. Fiederer, M. Glasstetter, K. Eggensperger, M. Tangermann, F. Hutter, W. Burgard, and T. Ball, “Deep learning with convolutional neural networks for EEG decoding and visualization,”Human Brain Mapping, vol. 38, no. 11, pp. 5391– 5420, 2017

    2017

Showing first 80 references.