Pith. sign in

REVIEW 3 major objections 2 minor 33 references

FDC-Net: Rethinking the association between EEG artifact removal and multi-dimensional affective computing

T0 review · 3 major / 2 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read FDC-Net couples EEG artifact removal with emotion recognition in one network

desk verdict The submission is two unrelated papers stapled together: the abstract promises an EEG denoising+emotion recognition network, the body is a chiral metamaterials chemistry study—so the claimed results have zero supporting evidence. read the letter →

arxiv 2508.05231 v2 pith:VZTXJR3Z submitted 2025-08-07 cs.HC cs.AI

Wenjia Dong , Xueyuan Xu , Tianze Yu , Junming Zhang , Li Zhuo This is my paper · ORCID
classification cs.HCcs.AI
keywords EEGemotionrecognitionartifactremovaljointoptimizationfeedback-drivennetworkfrequency-adaptiveTransformerDEAPDREAMERend-to-endlearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes FDC-Net, a network that removes physiological artifacts from EEG and recognizes emotions in the same pass instead of cleaning first and classifying later. It claims that feeding the emotion-recognition error back into the artifact-removal stage, while the denoising loss shapes the same pathway, avoids error accumulation and works even when artifacts remain. On the DEAP and DREAMER datasets it reports higher denoising correlation and higher recognition accuracy than nine baselines. A sympathetic reader would care because EEG-based emotion recognition is fragile in realistic settings and a single jointly trained model could remove a common failure point.

What carries the argument

The load-bearing object is FDC-Net's feedback-driven collaborative loop: error from the emotion-recognition head back-propagates into the artifact-removal stage through a shared representation, while the denoising loss bounds the same pathway. A gated attention mechanism inside a frequency-adaptive Transformer, with learnable band-position encoding, decides which EEG frequency bands the model relies on at each moment. This bidirectional coupling is what is supposed to let artifact removal and emotion recognition reinforce one another.

What would settle it

Locate the actual FDC-Net manuscript, implement the network exactly as it specifies, and rerun the DEAP and DREAMER evaluations; if the reported maximum correlation coefficients (96.30% and 90.31%) or recognition accuracies (82.3$\pm$7.1% and 88.1$\pm$0.8%) do not reproduce under that specified protocol, the central claim fails. Since the provided text does not contain this protocol, the first check is whether that description exists at all.

Watch

Extended reading notes

Core claim

On the paper's own account, FDC-Net establishes that EEG artifact removal and emotion recognition can be jointly optimized through bidirectional gradient propagation, so that the classifier's error sharpens the denoiser and the denoiser's constraints improve recognition. The network replaces the usual cascaded denoise-then-classify pipeline, and a gated attention mechanism inside a frequency-adaptive Transformer with learnable band-position encoding selects which frequency bands to trust. The reported figures are a maximum correlation coefficient of 96.30% on DEAP and 90.31% on DREAMER for denoising, and recognition accuracies of 82.3$\pm$7.1% on DEAP and 88.1$\pm$0.8% on DREAMER under physi

Load-bearing premise

The whole claim rests on the supplied text being FDC-Net's description of its architecture and experiments; in fact the full text is an unrelated paper about chiral metamaterials, so there is no method or protocol in front of the reader against which the reported accuracy numbers can be checked.

Editorial extensions

If this is right

  • Joint training of artifact removal and emotion recognition can outperform cascaded denoise-then-classify pipelines, reducing error accumulation.
  • The model can recognize emotions from EEG even when artifacts remain, weakening the idealized assumption that downstream analysis needs perfectly denoised data.
  • On DEAP and DREAMER the same architecture reports strong results for both tasks, suggesting one network can serve both signal cleaning and emotion decoding.

Reading between the lines

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

  • Because the supplied full text is a different study on chiral metamaterials, the abstract's numerical claims cannot currently be traced to an architecture description, training schedule, or baseline comparison; an independent replication would need the actual FDC-Net manuscript.
  • The same feedback-coupled design could be tried on other physiological signals such as ECG or EMG, where artifact removal and downstream classification are also staged separately; the paper does not report such results.
  • The reported accuracies use a mean-plus-standard-deviation format without specifying folds or trial counts in the supplied material, so the uncertainty around the headline numbers remains undefined at this level of description.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 2 minor

Summary. The paper, as submitted, claims to introduce FDC-Net, a Feedback-Driven Collaborative Network that couples EEG artifact removal and emotion recognition via bidirectional gradient propagation and a gated attention frequency-adaptive Transformer. The abstract reports specific numerical results on DEAP and DREAMER: maximum correlation coefficients of 96.30% and 90.31% for denoising, and emotion recognition accuracies of 82.3+7.1% and 88.1+0.8%, respectively, with comparisons against nine state-of-the-art methods. However, the full text provided is an entirely unrelated manuscript on chiral metamaterials, meta-diastereomers, and Mueller matrix polarimetry. No architecture details, equations, training protocols, dataset splits, baseline implementations, experimental results, or analyses for FDC-Net appear anywhere in the body. The central claim of the paper is therefore completely unsupported by the submitted manuscript.

Significance. If FDC-Net actually exists and performs as claimed, the idea of jointly optimizing artifact removal and emotion recognition through bidirectional gradient propagation and frequency-adaptive attention would be a potentially valuable contribution to EEG-based affective computing. The abstract's specificity suggests concrete performance improvements over cascaded approaches. However, the submitted full text contains none of the supporting material needed to evaluate the method's novelty, correctness, or validity. The significance assessment cannot go beyond the abstract because the body is an unrelated chemistry manuscript. No strengths related to reproducibility, formal proofs, or parameter-free derivations can be identified for FDC-Net from the provided text.

major comments (3)
  1. [Full Text (all sections)] The manuscript body is an unrelated paper about chiral metamaterials, meta-diastereomers, optical activity, and antibody-antigen detection. There is no mention of FDC-Net, EEG, DEAP, DREAMER, artifact removal, emotion recognition, or any of the claimed baselines. The central claim of the paper is absent from the submitted content. This is not a fixable gap in a section or equation; the entire supporting evidence is missing. The manuscript cannot be scientifically reviewed in its current form.
  2. [Abstract, performance claims] The abstract reports specific quantitative results: 96.30% CC on DEAP, 90.31% CC on DREAMER, 82.3+7.1% accuracy on DEAP, and 88.1+0.8% accuracy on DREAMER. None of these numbers can be traced to any experimental protocol, dataset split, evaluation metric definition, or statistical analysis in the full text. There is no way to verify whether these figures are from a held-out test set, whether they are averaged over subjects, or whether they represent statistically meaningful improvements. The notation '82.3+7.1' and '88.1+0.8' is also ambiguous (presumably standard deviations, but formatted with a plus sign).
  3. [Full Text (methods/comparison)] The abstract claims a comparison with nine state-of-the-art methods on DEAP and DREAMER. The full text provides no description of the baselines, no implementation details, no hyperparameter settings, no preprocessing pipeline for EEG artifact removal, and no tables or figures of comparative performance. Consequently, the central comparative claim—that joint optimization outperforms cascaded or independent approaches—cannot be assessed. The manuscript's title, abstract, and body are internally inconsistent, making any evaluation of the stated contribution impossible.
minor comments (2)
  1. [Abstract] The use of '82.3+7.1%' and '88.1+0.8%' should be '82.3 ± 7.1%' and '88.1 ± 0.8%' if these are standard deviations.
  2. [Full Text (title/author metadata)] The full text carries an unrelated title, author list, and keywords (meta-diastereomers, Mueller matrix polarimetry) that have no connection to the FDC-Net abstract. This indicates a severe submission error or a corrupted manuscript assembly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity can be identified because the supplied full text is an unrelated chiral-metamaterials manuscript; the FDC-Net derivation chain is absent.

full rationale

The abstract describes FDC-Net, a coupled EEG artifact-removal and emotion-recognition network, and reports CC/accuracy values on DEAP and DREAMER. However, the full text provided is a chemistry/nanophotonics paper titled 'Meta-Diastereomers Hierarchical Multiscale Chiral Interactions Between Biomolecules and Nanoscale Enantiomers.' It contains no mention of FDC-Net, EEG, DEAP, DREAMER, the nine baselines, the bidirectional gradient mechanism, the gated attention frequency-adaptive Transformer, or any training/evaluation protocol. Consequently there is no equation, fit, self-citation chain, or derivation in the manuscript that could reduce the reported results to their inputs. The reported performance figures are unsupported assertions, and the manuscript-body mismatch is a serious completeness/integrity problem, but it is not an instance of circular reasoning under the specified patterns. No circular step can be quoted because none is present; the score is therefore 0.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The abstract alone provides only high-level architectural components, leading to a small set of free parameters and domain assumptions. The full text mismatch prevents a more detailed accounting.

free parameters (2)
  • learnable band-position encoding
    The abstract mentions learnable band-position encoding in the frequency-adaptive Transformer, which implies learned parameters in the position encoder.
  • gated attention parameters
    The gated attention mechanism includes learned gates and attention weights that are optimized during training.
assumptions (3)
  • domain assumption DEAP and DREAMER provide reliable ground-truth emotion labels for supervised training and evaluation.
    The abstract relies on these datasets for both denoising and emotion recognition metrics, but no details on label validity or preprocessing are given.
  • domain assumption Physiological artifacts can be systematically removed from EEG without losing emotion-relevant information.
    The central premise of the framework is that denoising is beneficial for emotion recognition and that joint training improves both tasks.
  • domain assumption Bidirectional gradient propagation between denoising and classification yields synergy.
    The abstract asserts this as the primary innovation, but no theoretical or empirical support is provided in the text.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FDC-Net: Rethinking the association between EEG artifact removal and multi-dimensional affective computing." pith.science (2026). https://pith.science/paper/VZTXJR3Z

@misc{pith2026250805231,
  author       = {Pith},
  title        = {Pith review of: FDC-Net: Rethinking the association between EEG artifact removal and multi-dimensional affective computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VZTXJR3Z}},
  note         = {Machine review of arXiv:2508.05231}
}
read the original abstract

Electroencephalogram (EEG)-based emotion recognition holds significant value in affective computing and brain-computer interfaces. However, in practical applications, EEG recordings are susceptible to the effects of various physiological artifacts. Current approaches typically treat denoising and emotion recognition as independent tasks using cascaded architectures, which not only leads to error accumulation, but also fails to exploit potential synergies between these tasks. Moreover, conventional EEG-based emotion recognition models often rely on the idealized assumption of "perfectly denoised data", lacking a systematic design for noise robustness. To address these challenges, a novel framework that deeply couples denoising and emotion recognition tasks is proposed for end-to-end noise-robust emotion recognition, termed as Feedback-Driven Collaborative Network for Denoising-Classification Nexus (FDC-Net). Our primary innovation lies in establishing a dynamic collaborative mechanism between artifact removal and emotion recognition through: (1) bidirectional gradient propagation with joint optimization strategies; (2) a gated attention mechanism integrated with frequency-adaptive Transformer using learnable band-position encoding. Two most popular EEG-based emotion datasets (DEAP and DREAMER) with multi-dimensional emotional labels were employed to compare the artifact removal and emotion recognition performance between FDC-Net and nine state-of-the-art methods. In terms of the denoising task, FDC-Net obtains a maximum correlation coefficient (CC) value of 96.30% on DEAP and a maximum CC value of 90.31% on DREAMER. In terms of the emotion recognition task under physiological artifact interference, FDC-Net achieves emotion recognition accuracies of 82.3+7.1% on DEAP and 88.1+0.8% on DREAMER.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 31 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Adeli, H.; Zhou, Z.; and Dadmehr, N. 2003. Analysis of EEG records in an epileptic patient using wavelet transform. Journal of neuroscience methods, 123(1): 69--87

    work page 2003

  4. [4]

    M.; Hassan, M

    Al-Selwi, S. M.; Hassan, M. F.; Abdulkadir, S. J.; Muneer, A.; Sumiea, E. H.; Alqushaibi, A.; and Ragab, M. G. 2024. RNN-LSTM: From applications to modeling techniques and beyond—Systematic review. Journal of King Saud University-Computer and Information Sciences, 36(5): 102068

    work page 2024

  5. [5]

    Albera, L.; Kachenoura, A.; Comon, P.; Karfoul, A.; Wendling, F.; Senhadji, L.; and Merlet, I. 2012. ICA-based EEG denoising: a comparative analysis of fifteen methods. Bulletin of the Polish Academy of Sciences: Technical Sciences, 60(3 Special issue on Data Mining in Bioengineering): 407--418

    work page 2012

  6. [6]

    Cai, Y.; Meng, Z.; and Huang, D. 2025. DHCT-GAN: Improving EEG signal quality with a dual-branch hybrid CNN--transformer network. Sensors, 25(1): 231

    work page 2025

  7. [7]

    Chen, J.; Pi, D.; Jiang, X.; Xu, Y.; Chen, Y.; and Wang, X. 2023. Denosieformer: A transformer-based approach for single-channel EEG artifact removal. IEEE Transactions on Instrumentation and Measurement, 73: 1--16

    work page 2023

  8. [8]

    Chuang, C.-H.; Chang, K.-Y.; Huang, C.-S.; and Bessas, A.-M. 2025. Augmenting brain-computer interfaces with ART: An artifact removal transformer for reconstructing multichannel EEG signals. NeuroImage, 310: 121123

    work page 2025

Show all 33 references
  1. [9]

    Cui, H.; Li, C.; Liu, A.; Qian, R.; and Chen, X. 2024. A Dual-Branch Interactive Fusion Network to Remove Artifacts From Single-Channel EEG. IEEE Transactions on Instrumentation and Measurement, 73: 1--12

    2024

  2. [10]

    Davidson, R. J. 2003. Affective neuroscience and psychophysiology: Toward a synthesis. Psychophysiology, 40(5): 655--665

    2003

  3. [11]

    S.; and Rosa, J

    de Aguiar Neto, F. S.; and Rosa, J. L. G. 2019. Depression biomarkers using non-invasive EEG: A review. Neuroscience & Biobehavioral Reviews, 105: 83--93

    2019

  4. [12]

    J.; and Guan, C

    Ding, Y.; Tong, C.; Zhang, S.; Jiang, M.; Li, Y.; Lim, K. J.; and Guan, C. 2025. EmT: A novel transformer for generalized cross-subject EEG emotion recognition. IEEE Transactions on Neural Networks and Learning Systems

    2025

  5. [13]

    Frey, J.; M \"u hl, C.; Lotte, F.; and Hachet, M. 2013. Review of the use of electroencephalography as an evaluation method for human-computer interaction. arXiv preprint arXiv:1311.2222

    2013 arXiv

  6. [14]

    Gamage, T.; Kalansooriya, L.; and Sandamali, E. 2022. An emotion classification model for driver emotion recognition using electroencephalography (EEG). In 2022 international research conference on smart computing and systems engineering (SCSE), volume 5, 76--82. IEEE

    2022

  7. [15]

    Kendall, A.; Gal, Y.; and Cipolla, R. 2018. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In Proceedings of the IEEE conference on computer vision and pattern recognition, 7482--7491

    2018

  8. [16]

    Kher, R.; and Gandhi, R. 2016. Adaptive filtering based artifact removal from electroencephalogram (EEG) signals. In 2016 International Conference on Communication and Signal Processing (ICCSP), 0561--0564. IEEE

    2016

  9. [17]

    Q.; Ibrahim, H.; Abdullah, M

    Lai, C. Q.; Ibrahim, H.; Abdullah, M. Z.; Abdullah, J. M.; Suandi, S. A.; and Azman, A. 2018. Artifacts and noise removal for electroencephalogram (EEG): A literature review. In 2018 IEEE Symposium on Computer Applications & Industrial Electronics (ISCAIE), 326--332

    2018

  10. [18]

    Lu, W.; Ma, H.; and Tan, T.-P. 2023. CIT-EmotionNet: CNN interactive transformer network for EEG emotion recognition. arXiv preprint arXiv:2305.05548

    2023 arXiv

  11. [19]

    Paulhus, D. L. 1991. Measurement and control of response bias

    1991

  12. [20]

    Pratt, H.; Williams, B.; Coenen, F.; and Zheng, Y. 2017. Fcnn: Fourier convolutional neural networks. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, 786--798. Springer

    2017

  13. [21]

    N.; Rosenkranz, T.; and Maier, A

    Schroter, H.; Escalante-B, A. N.; Rosenkranz, T.; and Maier, A. 2022. DeepFilterNet: A low complexity speech enhancement framework for full-band audio based on deep filtering. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)...

    2022

  14. [22]

    Shad, E. H. T.; Molinas, M.; and Ytterdal, T. 2020. Impedance and noise of passive and active dry EEG electrodes: a review. IEEE Sensors Journal, 20(24): 14565--14577

    2020

  15. [23]

    Song, T.; Zheng, W.; Song, P.; and Cui, Z. 2018. EEG emotion recognition using dynamical graph convolutional neural networks. IEEE Transactions on Affective Computing, 11(3): 532--541

    2018

  16. [24]

    Sun, W.; Su, Y.; Wu, X.; and Wu, X. 2020. A novel end-to-end 1D-ResCNN model to remove artifact from EEG signals. Neurocomputing, 404: 108--121

    2020

  17. [25]

    Tao, W.; Li, C.; Song, R.; Cheng, J.; Liu, Y.; Wan, F.; and Chen, X. 2020. EEG-based emotion recognition via channel-wise attention and self attention. IEEE Transactions on Affective Computing, 14(1): 382--393

    2020

  18. [26]

    Wang, Y.; Peng, Y.; Han, M.; Liu, X.; Niu, H.; Cheng, J.; Chang, S.; and Liu, T. 2024. GCTNet: a graph convolutional transformer network for major depressive disorder detection based on EEG signals. Journal of Neural Engineering, 21(3): 036042

    2024

  19. [27]

    Wang, Y.; Song, W.; Tao, W.; Liotta, A.; Yang, D.; Li, X.; Gao, S.; Sun, Y.; Ge, W.; Zhang, W.; et al. 2022. A systematic review on affective computing: Emotion models, databases, and recent advances. Information Fusion, 83: 19--52

    2022

  20. [28]

    Wang, Y.; Zhang, B.; and Tang, Y. 2024. DMMR: Cross-subject domain generalization for EEG-based emotion recognition via denoising mixed mutual reconstruction. In Proceedings of the AAAI conference on artificial intelligence, volume 38, 628--636

    2024

  21. [29]

    Yang, D.; Liu, Y.; Zhou, Z.; Yu, Y.; and Liang, X. 2020. Decoding visual motions from EEG using attention-based RNN. Applied Sciences, 10(16): 5662

    2020

  22. [30]

    Yao, X.; Li, T.; Ding, P.; Wang, F.; Zhao, L.; Gong, A.; Nan, W.; and Fu, Y. 2024. Emotion classification based on transformer and CNN for EEG spatial--temporal feature learning. Brain sciences, 14(3): 268

    2024

  23. [31]

    P.; and Zhang, T

    Ye, M.; Chen, C. P.; and Zhang, T. 2022. Hierarchical dynamic graph convolutional network with interpretability for EEG-based emotion recognition. IEEE transactions on neural networks and learning systems

    2022

  24. [32]

    Zamani, F.; Wulansari, R.; et al. 2021. Emotion classification using 1D-CNN and RNN based on deap dataset. Nat. Lang. Process, 363--378

    2021

  25. [33]

    Zhang, Z.; Zhong, S.; and Liu, Y. 2024. Beyond mimicking under-represented emotions: Deep data augmentation with emotional subspace constraints for EEG-based emotion recognition. In Proceedings of the AAAI conference on artificial intelligence, volume 38, 10252--10260

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.