Spectral Priors vs. Attention: Investigating the Utility of Attention Mechanisms in EEG-Based Diagnosis

Gowtham Atluri; Tawsik Jawad; Vikram Ravindra

arxiv: 2605.15433 · v2 · pith:DOJBNU2Znew · submitted 2026-05-14 · 💻 cs.LG

Spectral Priors vs. Attention: Investigating the Utility of Attention Mechanisms in EEG-Based Diagnosis

Tawsik Jawad , Gowtham Atluri , Vikram Ravindra This is my paper

Pith reviewed 2026-06-30 20:53 UTC · model grok-4.3

classification 💻 cs.LG

keywords EEGspectral featuresattention mechanismsmachine learningneurodegenerative diseasesclassificationbrainwave bandssmall datasets

0 comments

The pith

Spectral features from brainwave bands enable traditional machine learning to match or exceed deep learning performance on small EEG datasets, while attention mechanisms fail to extract stable healthy neural signatures.

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

EEG signals for diagnosing neurodegenerative diseases are noisy with coarse resolution, making classification difficult even for state-of-the-art deep learning models. The paper establishes that constructing features by isolating strengths in primary brainwave frequency bands creates high-value inputs that improve class separability. On small datasets, these frequency and time-frequency features allow conventional machine learning models to perform as well as or better than deep learning architectures. Attention-based models specifically cannot identify the consistent spectral signatures of healthy brain activity in resting or task-based recordings. The finding is validated on three resting-state and one task EEG dataset, indicating that spectral priors may be more effective than attention for limited data scenarios.

Core claim

By isolating signal strengths within the primary brainwave bands, high dimensional raw EEG data is transformed into high value spectral features. In small datasets, features derived from frequency and time-frequency domain allow traditional machine learning models to match or exceed the performance of SOTA deep learning models. Attention mechanisms are unable to distill the stable feature signatures that characterize healthy neural activity in both resting and task EEGs. The limitations of attention based models in finding relevant spectral features appear robust, as providing frequency selective time domain input does not appreciably improve their performance.

What carries the argument

Spectrally selective feature construction isolating signal strengths in primary brainwave bands to enhance class separability.

If this is right

Traditional machine learning models using frequency and time-frequency features match or exceed SOTA deep learning models in small datasets.
Attention mechanisms cannot distill stable feature signatures of healthy neural activity.
Limitations of attention models persist even when given frequency selective time domain inputs.
Findings hold across three open source resting EEG datasets and one task EEG dataset.

Where Pith is reading between the lines

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

Explicit incorporation of known frequency band structures may outperform learned attention when training data is scarce for noisy signals.
Testing the approach on larger EEG collections could determine if attention benefits emerge with more data.
Similar spectral feature strategies might apply to other time-series classification tasks with known frequency-domain priors, such as other biomedical signals.

Load-bearing premise

The observed performance gap stems from the spectral feature construction rather than differences in hyperparameter tuning, model capacity, or preprocessing between the traditional ML and attention pipelines.

What would settle it

Re-running the model comparisons after equalizing hyperparameter search effort, model parameter counts, and all preprocessing steps to isolate whether the spectral feature step alone accounts for the gap.

Figures

Figures reproduced from arXiv: 2605.15433 by Gowtham Atluri, Tawsik Jawad, Vikram Ravindra.

**Figure 1.** Figure 1: Holdout-set confusion matrices on ADFTD comparing a classical pipeline vs. a Transformer. Rows denote ground-truth labels (‘A’:Alzheimers, ‘C’:Healthy Controls, ‘F’:Dementia) and columns denote predicted labels; darker diagonal entries indicate better class-wise performance. Left: Quadratic Discriminant Analysis (QDA) trained on aggregated spectral features (Welch-FFT/DWT, channeland window-averaged) show… view at source ↗

read the original abstract

Electroencephalograph (EEG) timeseries signals are characterized by significant noise and coarse spatial resolution, which complicates the classification of neurodegenerative diseases. Even SOTA deep learning architectures struggle to distinguish between healthy controls and diseased subjects, or between different disease types, due to high intergroup similarity. In this paper, we show that a spectrally selective approach to feature construction enhances class separability. By isolating signal strengths within the primary brainwave bands, we transform high dimensional raw data into high value spectral features. Our results demonstrate that in small datasets a) features derived from frequency and time frequency domain allow traditional machine learning models to match or exceed the performance of SOTA deep learning models, b) Attention mechanism is unable to distill the stable feature signatures that characterize healthy neural activity in both resting and task EEGs, and c) the limitations of attention based models in finding relevant spectral features appear to be robust in that providing frequency selective time domain input do not appreciably improve their performance. We validate our methodology across three open source resting EEG datasets and one task EEG dataset, providing robust empirical evidence for our claims.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Spectral features with traditional ML match attention-based DL on small EEG datasets, but missing numbers and method details make the comparison hard to trust.

read the letter

The main thing to know is that this paper reports spectral and time-frequency features let classic ML match or beat attention-heavy deep models on four open EEG datasets for disease classification, and that attention still underperforms even after an ablation that feeds it frequency-selective time-domain inputs.

The work does a direct head-to-head on three resting-state and one task EEG dataset. That setup tests a practical claim: on the small samples typical in clinical EEG, explicit spectral priors may be more reliable than letting attention discover the same structure. The ablation is a reasonable check on whether the attention limitation is just about input format.

The soft spot is exactly the one flagged in the stress-test note. The abstract gives no performance numbers, no error bars, no statistical tests, and no information on hyperparameter search, regularization, or preprocessing differences between the ML and DL pipelines. Without those, it is impossible to separate the effect of the spectral features from unequal tuning effort or capacity. If the attention models received lighter optimization, the ordering could be an artifact.

This is for readers who build or evaluate EEG diagnostic tools on limited data and want a focused empirical check on when attention adds value. It is worth sending to peer review because the datasets are public and the question is actionable, but only after the authors add the actual results and implementation details so referees can assess whether the comparison is fair.

Referee Report

2 major / 0 minor

Summary. The paper claims that in small EEG datasets for neurodegenerative disease diagnosis, explicit spectral features derived from frequency and time-frequency domains enable traditional ML models to match or exceed SOTA deep learning models incorporating attention; that attention mechanisms cannot distill stable spectral signatures of healthy neural activity; and that frequency-selective time-domain inputs do not appreciably improve attention model performance. These claims are supported by empirical validation across three resting-state and one task-based open-source EEG dataset.

Significance. If the performance ordering holds under controlled conditions, the work would provide concrete evidence favoring domain-informed spectral feature construction over attention-based learning for EEG tasks with limited samples, with potential implications for model design in noisy, low-resolution biosignal classification. The multi-dataset validation is a positive aspect for generalizability.

major comments (2)

[Abstract / Experimental Setup] Abstract and Experimental Setup (inferred from results description): The manuscript reports that attention models were given frequency-selective time-domain inputs as an ablation, yet provides no information on whether these models received equivalent hyperparameter search effort, regularization strength, or optimization budget compared to the spectral-feature ML baselines. This detail is load-bearing for the central claim that attention is inherently unable to distill spectral signatures rather than the gap arising from unequal experimental conditions.
[Results] Results section: No quantitative performance metrics, error bars, statistical significance tests, or data-split details are referenced in the abstract or summary of findings, despite claims of matching/exceeding performance across four datasets. Without these, the magnitude and reliability of the reported gaps cannot be assessed.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments, which highlight important aspects of experimental rigor and reporting. We address each point below and will revise the manuscript accordingly to strengthen the presentation of our claims.

read point-by-point responses

Referee: [Abstract / Experimental Setup] The manuscript reports that attention models were given frequency-selective time-domain inputs as an ablation, yet provides no information on whether these models received equivalent hyperparameter search effort, regularization strength, or optimization budget compared to the spectral-feature ML baselines. This detail is load-bearing for the central claim that attention is inherently unable to distill spectral signatures rather than the gap arising from unequal experimental conditions.

Authors: We agree this information is necessary to support the interpretation that performance gaps reflect inherent limitations of attention rather than experimental imbalance. All models, including attention ablations, received grid searches over comparable hyperparameter spaces (learning rate, regularization, architecture depth/width, optimizer settings) with equivalent total optimization budget and early stopping criteria. To make this explicit, we will add a new subsection under Experimental Setup that tabulates the search ranges, selected values, and validation procedure for every model family and ablation variant. revision: yes
Referee: [Results] No quantitative performance metrics, error bars, statistical significance tests, or data-split details are referenced in the abstract or summary of findings, despite claims of matching/exceeding performance across four datasets. Without these, the magnitude and reliability of the reported gaps cannot be assessed.

Authors: The full manuscript already contains tables reporting mean accuracy/F1, standard deviations from 5-fold subject-independent cross-validation, and paired statistical tests (p-values) for all four datasets, along with explicit train/val/test split ratios. These details were omitted from the abstract for brevity. We will revise the abstract to include representative quantitative results (e.g., accuracy ranges and significance statements) and will add a short results-summary paragraph that explicitly references the error bars and statistical tests already present in the tables. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical comparison with no derivations or self-referential predictions

full rationale

The paper is an empirical study comparing spectral-feature-based traditional ML against attention-based DL models on four open EEG datasets. The abstract and available text contain no equations, first-principles derivations, fitted parameters renamed as predictions, or load-bearing self-citations. All claims rest on reported classification accuracies rather than any chain that reduces to its own inputs by construction. This matches the default case of a self-contained empirical paper against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper relies on standard assumptions of supervised classification and the existence of distinct spectral signatures in EEG bands; no free parameters, invented entities, or ad-hoc axioms are introduced in the abstract.

axioms (1)

domain assumption EEG signals contain stable, class-discriminative information in standard frequency bands that can be isolated without loss of diagnostic utility.
Invoked when the paper states that isolating signal strengths within primary brainwave bands transforms raw data into high-value spectral features.

pith-pipeline@v0.9.1-grok · 5729 in / 1439 out tokens · 25902 ms · 2026-06-30T20:53:52.731109+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

27 extracted references · 7 canonical work pages · 4 internal anchors

[1]

Simplified welch algorithm for spectrum monitoring

Same, Mohammad Hossein, et al. "Simplified welch algorithm for spectrum monitoring." Applied Sciences 11.1 (2020): 86

2020
[2]

Nearest neighbors in high-dimensional data: The emergence and influence of hubs

Radovanović, Miloš, Alexandros Nanopoulos, and Mirjana Ivanović. “Nearest neighbors in high-dimensional data: The emergence and influence of hubs." Proceedings of the 26th Annual International Conference on Machine Learning. 2009

2009
[3]

Linear and Quadratic Discriminant Analysis: Tutorial

Ghojogh, B., and M. Crowley. “Linear and quadratic discriminant analysis: Tutorial. arXiv 2019." arXiv preprint arXiv:1906.02590

work page internal anchor Pith review Pith/arXiv arXiv 2019
[4]

Medformer: A multi-granularity patching transformer for medical time-series classification

Wang, Yihe, et al. "Medformer: A multi-granularity patching transformer for medical time-series classification." Advances in Neural Information Processing Systems 37 (2024): 36314-36341

2024
[5]

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

Wu, Haixu, et al. “Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting.”NeurIPS34 (2021): 22419–22430

2021
[6]

STL: A seasonal-trend decomposition

Cleveland, Robert B., et al. "STL: A seasonal-trend decomposition." J. off. Stat 6.1 (1990): 3-73. 10 Tawsik Jawad, Gowtham Atluri, and Vikram Ravindra

1990
[7]

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

Zhou, Tian, et al. "Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting." International conference on machine learning. PMLR, 2022

2022
[8]

Reformer: The Efficient Transformer

Kitaev, Nikita, Łukasz Kaiser, and Anselm Levskaya. "Reformer: The efficient transformer." arXiv preprint arXiv:2001.04451 (2020)

work page internal anchor Pith review Pith/arXiv arXiv 2001
[9]

Analysis of electroencephalograms in alzheimer’s disease patients with multiscale entropy

J Escudero et al. Analysis of electroencephalograms in alzheimer’s disease patients with multiscale entropy. Physiological measurement, 27(11):1091, 2006

2006
[10]

The two decades brainclinics research archive for insights in neurophysiology (tdbrain) database

Hanneke van Dijk, Guido van Wingen, Damiaan Denys, Sebastian Olbrich, Rosalinde van Ruth, and Martijn Arns. The two decades brainclinics research archive for insights in neurophysiology (tdbrain) database. Scientific data, 9(1):333, 2022

2022
[11]

A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg

Andreas Miltiadous, Katerina D Tzimourta et al. A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg. Data, 8(6):95, 2023

2023
[12]

Windowing Techniques, the welch method for improvement of Power Spectrum Estimation

Jwo, Dah-Jing, Wei-Yeh Chang, and I-Hua Wu. "Windowing Techniques, the welch method for improvement of Power Spectrum Estimation." Computers, materials & continua 67.3 (2021)

2021
[13]

Intra- and Inter-subject Variability in EEG-Based Sensorimotor Brain Computer Interface: A Review

Saha S, Baumert M. Intra- and Inter-subject Variability in EEG-Based Sensorimotor Brain Computer Interface: A Review. Front Comput Neurosci. 2020 Jan 21;13:87. doi: 10.3389/fncom.2019.00087. PMID: 32038208; PMCID: PMC6985367

work page doi:10.3389/fncom.2019.00087 2020
[14]

Selective cross-subject transfer learning based on riemannian tangent space for motor imagery brain-computer interface

Xu, Yilu, Xin Huang, and Quan Lan. "Selective cross-subject transfer learning based on riemannian tangent space for motor imagery brain-computer interface." Frontiers in Neuroscience 15 (2021): 779231

2021
[15]

Cross-dataset variability problem in EEG decoding with deep learning

Xu, Lichao, et al. "Cross-dataset variability problem in EEG decoding with deep learning." Frontiers in human neuroscience 14 (2020): 103

2020
[16]

Adaptive deep feature representation learning for cross-subject EEG decoding

Liang, Shuang, et al. "Adaptive deep feature representation learning for cross-subject EEG decoding." BMC bioinformatics 25.1 (2024): 393

2024
[17]

ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels

Dempster, Angus, François Petitjean, and Geoffrey I. Webb. "ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels." Data Mining and Knowledge Discovery 34.5 (2020): 1454-1495

2020
[18]

TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis

Wu, Haixu, et al. "Timesnet: Temporal 2d-variation modeling for general time series analysis." arXiv preprint arXiv:2210.02186 (2022)

work page internal anchor Pith review Pith/arXiv arXiv 2022
[19]

Systematic review on resting-state EEG for Alzheimer’s disease diagnosis and progression assessment

Cassani, Raymundo, et al. "Systematic review on resting-state EEG for Alzheimer’s disease diagnosis and progression assessment." Disease markers 2018.1 (2018): 5174815

2018
[21]

A dementia classification framework using frequency and time-frequency features based on EEG signals

Durongbhan, Pholpat, et al. "A dementia classification framework using frequency and time-frequency features based on EEG signals." IEEE Transactions on Neural Systems and Rehabilitation Engineering 27.5 (2019): 826-835

2019
[22]

EEG Conformer: Convolutional Transformer for EEG Decoding and Visualization,

Y. Song, Q. Zheng, B. Liu and X. Gao, "EEG Conformer: Convolutional Transformer for EEG Decoding and Visualization," in IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 31, pp. 710-719, 2023, doi: 10.1109/TNSRE.2022.3230250

work page doi:10.1109/tnsre.2022.3230250 2023
[23]

Classification of EEG signals using Transformer based deep learning and ensemble models

Zeynali, Mahsa, Hadi Seyedarabi, and Reza Afrouzian. "Classification of EEG signals using Transformer based deep learning and ensemble models." Biomedical Signal Processing and Control 86 (2023): 105130

2023
[24]

A Hybrid Approach to Attention Deficit Hyperactivity Disorder Detection Leveraging Transformer and XGBoost Models Using XSparseFormerNet

Sarker SR et al. A Hybrid Approach to Attention Deficit Hyperactivity Disorder Detection Leveraging Transformer and XGBoost Models Using XSparseFormerNet. Sci Rep. 2025 Nov 20;15(1):41039. doi: 10.1038/s41598-025-24919-3. PMID: 41266583; PMCID: PMC12635201

work page doi:10.1038/s41598-025-24919-3 2025
[25]

Gamma oscillations in the hyperkinetic state detected with chronic human brain recordings in Parkinson’s disease

Swann, Nicole C., et al. "Gamma oscillations in the hyperkinetic state detected with chronic human brain recordings in Parkinson’s disease." Journal of Neuroscience 36.24 (2016): 6445-6458. Spectral Priors vs Self Attention 11

2016
[26]

High delta and gamma EEG power in resting state characterise dementia in Parkinson’s patients

Pal, Anita, et al. "High delta and gamma EEG power in resting state characterise dementia in Parkinson’s patients." Biomarkers in Neuropsychiatry 3 (2020): 100027

2020
[27]

EEG data for ADHD / Control children

Ali Motie Nasrabadi, Armin Allahverdy, Mehdi Samavati, Mohammad Reza Mohammadi.(2020). "EEG data for ADHD / Control children." Web,

2020
[28]

Decentralized Attention Fails Centralized Signals: Rethinking Transformers for Medical Time Series

Yu, Guoqi, et al. "Decentralized Attention Fails Centralized Signals: Rethinking Transformers for Medical Time Series." arXiv preprint arXiv:2602.18473 (2026)

work page internal anchor Pith review Pith/arXiv arXiv 2026

[1] [1]

Simplified welch algorithm for spectrum monitoring

Same, Mohammad Hossein, et al. "Simplified welch algorithm for spectrum monitoring." Applied Sciences 11.1 (2020): 86

2020

[2] [2]

Nearest neighbors in high-dimensional data: The emergence and influence of hubs

Radovanović, Miloš, Alexandros Nanopoulos, and Mirjana Ivanović. “Nearest neighbors in high-dimensional data: The emergence and influence of hubs." Proceedings of the 26th Annual International Conference on Machine Learning. 2009

2009

[3] [3]

Linear and Quadratic Discriminant Analysis: Tutorial

Ghojogh, B., and M. Crowley. “Linear and quadratic discriminant analysis: Tutorial. arXiv 2019." arXiv preprint arXiv:1906.02590

work page internal anchor Pith review Pith/arXiv arXiv 2019

[4] [4]

Medformer: A multi-granularity patching transformer for medical time-series classification

Wang, Yihe, et al. "Medformer: A multi-granularity patching transformer for medical time-series classification." Advances in Neural Information Processing Systems 37 (2024): 36314-36341

2024

[5] [5]

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

Wu, Haixu, et al. “Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting.”NeurIPS34 (2021): 22419–22430

2021

[6] [6]

STL: A seasonal-trend decomposition

Cleveland, Robert B., et al. "STL: A seasonal-trend decomposition." J. off. Stat 6.1 (1990): 3-73. 10 Tawsik Jawad, Gowtham Atluri, and Vikram Ravindra

1990

[7] [7]

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

Zhou, Tian, et al. "Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting." International conference on machine learning. PMLR, 2022

2022

[8] [8]

Reformer: The Efficient Transformer

Kitaev, Nikita, Łukasz Kaiser, and Anselm Levskaya. "Reformer: The efficient transformer." arXiv preprint arXiv:2001.04451 (2020)

work page internal anchor Pith review Pith/arXiv arXiv 2001

[9] [9]

Analysis of electroencephalograms in alzheimer’s disease patients with multiscale entropy

J Escudero et al. Analysis of electroencephalograms in alzheimer’s disease patients with multiscale entropy. Physiological measurement, 27(11):1091, 2006

2006

[10] [10]

The two decades brainclinics research archive for insights in neurophysiology (tdbrain) database

Hanneke van Dijk, Guido van Wingen, Damiaan Denys, Sebastian Olbrich, Rosalinde van Ruth, and Martijn Arns. The two decades brainclinics research archive for insights in neurophysiology (tdbrain) database. Scientific data, 9(1):333, 2022

2022

[11] [11]

A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg

Andreas Miltiadous, Katerina D Tzimourta et al. A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg. Data, 8(6):95, 2023

2023

[12] [12]

Windowing Techniques, the welch method for improvement of Power Spectrum Estimation

Jwo, Dah-Jing, Wei-Yeh Chang, and I-Hua Wu. "Windowing Techniques, the welch method for improvement of Power Spectrum Estimation." Computers, materials & continua 67.3 (2021)

2021

[13] [13]

Intra- and Inter-subject Variability in EEG-Based Sensorimotor Brain Computer Interface: A Review

Saha S, Baumert M. Intra- and Inter-subject Variability in EEG-Based Sensorimotor Brain Computer Interface: A Review. Front Comput Neurosci. 2020 Jan 21;13:87. doi: 10.3389/fncom.2019.00087. PMID: 32038208; PMCID: PMC6985367

work page doi:10.3389/fncom.2019.00087 2020

[14] [14]

Selective cross-subject transfer learning based on riemannian tangent space for motor imagery brain-computer interface

Xu, Yilu, Xin Huang, and Quan Lan. "Selective cross-subject transfer learning based on riemannian tangent space for motor imagery brain-computer interface." Frontiers in Neuroscience 15 (2021): 779231

2021

[15] [15]

Cross-dataset variability problem in EEG decoding with deep learning

Xu, Lichao, et al. "Cross-dataset variability problem in EEG decoding with deep learning." Frontiers in human neuroscience 14 (2020): 103

2020

[16] [16]

Adaptive deep feature representation learning for cross-subject EEG decoding

Liang, Shuang, et al. "Adaptive deep feature representation learning for cross-subject EEG decoding." BMC bioinformatics 25.1 (2024): 393

2024

[17] [17]

ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels

Dempster, Angus, François Petitjean, and Geoffrey I. Webb. "ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels." Data Mining and Knowledge Discovery 34.5 (2020): 1454-1495

2020

[18] [18]

TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis

Wu, Haixu, et al. "Timesnet: Temporal 2d-variation modeling for general time series analysis." arXiv preprint arXiv:2210.02186 (2022)

work page internal anchor Pith review Pith/arXiv arXiv 2022

[19] [19]

Systematic review on resting-state EEG for Alzheimer’s disease diagnosis and progression assessment

Cassani, Raymundo, et al. "Systematic review on resting-state EEG for Alzheimer’s disease diagnosis and progression assessment." Disease markers 2018.1 (2018): 5174815

2018

[20] [21]

A dementia classification framework using frequency and time-frequency features based on EEG signals

Durongbhan, Pholpat, et al. "A dementia classification framework using frequency and time-frequency features based on EEG signals." IEEE Transactions on Neural Systems and Rehabilitation Engineering 27.5 (2019): 826-835

2019

[21] [22]

EEG Conformer: Convolutional Transformer for EEG Decoding and Visualization,

Y. Song, Q. Zheng, B. Liu and X. Gao, "EEG Conformer: Convolutional Transformer for EEG Decoding and Visualization," in IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 31, pp. 710-719, 2023, doi: 10.1109/TNSRE.2022.3230250

work page doi:10.1109/tnsre.2022.3230250 2023

[22] [23]

Classification of EEG signals using Transformer based deep learning and ensemble models

Zeynali, Mahsa, Hadi Seyedarabi, and Reza Afrouzian. "Classification of EEG signals using Transformer based deep learning and ensemble models." Biomedical Signal Processing and Control 86 (2023): 105130

2023

[23] [24]

A Hybrid Approach to Attention Deficit Hyperactivity Disorder Detection Leveraging Transformer and XGBoost Models Using XSparseFormerNet

Sarker SR et al. A Hybrid Approach to Attention Deficit Hyperactivity Disorder Detection Leveraging Transformer and XGBoost Models Using XSparseFormerNet. Sci Rep. 2025 Nov 20;15(1):41039. doi: 10.1038/s41598-025-24919-3. PMID: 41266583; PMCID: PMC12635201

work page doi:10.1038/s41598-025-24919-3 2025

[24] [25]

Gamma oscillations in the hyperkinetic state detected with chronic human brain recordings in Parkinson’s disease

Swann, Nicole C., et al. "Gamma oscillations in the hyperkinetic state detected with chronic human brain recordings in Parkinson’s disease." Journal of Neuroscience 36.24 (2016): 6445-6458. Spectral Priors vs Self Attention 11

2016

[25] [26]

High delta and gamma EEG power in resting state characterise dementia in Parkinson’s patients

Pal, Anita, et al. "High delta and gamma EEG power in resting state characterise dementia in Parkinson’s patients." Biomarkers in Neuropsychiatry 3 (2020): 100027

2020

[26] [27]

EEG data for ADHD / Control children

Ali Motie Nasrabadi, Armin Allahverdy, Mehdi Samavati, Mohammad Reza Mohammadi.(2020). "EEG data for ADHD / Control children." Web,

2020

[27] [28]

Decentralized Attention Fails Centralized Signals: Rethinking Transformers for Medical Time Series

Yu, Guoqi, et al. "Decentralized Attention Fails Centralized Signals: Rethinking Transformers for Medical Time Series." arXiv preprint arXiv:2602.18473 (2026)

work page internal anchor Pith review Pith/arXiv arXiv 2026