T-TIME: Test-Time Information Maximization Ensemble for Plug-and-Play BCIs

Dongrui Wu; Hanbin Luo; Lieyun Ding; Siyang Li; Ziwei Wang

REVIEW 3 major objections 6 minor 44 references

T-TIME: Test-Time Information Maximization Ensemble for Plug-and-Play BCIs

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Test-time adaptation can eliminate EEG BCI calibration while matching offline accuracy.

desk verdict T-TIME is a solid, reproducible first step toward calibration-free BCIs, but the spectral meta-learner extension is asserted without proof and needs to be fixed before acceptance. read the letter →

arxiv 2412.07228 v1 pith:73DQ4MMY submitted 2024-12-10 cs.HC cs.LG

Siyang Li , Ziwei Wang , Hanbin Luo , Lieyun Ding , Dongrui Wu This is my paper · ORCID

classification cs.HCcs.LG

keywords test-timeadaptationEEGbrain-computerinterfacemotorimagerytransferlearningensemblecalibration-freeBCIclassimbalance

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper aims to establish that test-time adaptation (TTA) can remove the need for a calibration session in EEG-based brain-computer interfaces, even in the hardest online setting where the new user's trials arrive one at a time and must be classified immediately. The proposed method, T-TIME, trains several EEGNet classifiers on existing users' data aligned by Euclidean alignment, predicts each incoming unlabeled trial by a spectral meta-learner combination of their outputs, and then updates every classifier with a loss that combines conditional entropy minimization and an adaptive marginal label distribution regularizer. Across three public motor-imagery datasets, the paper reports that T-TIME outperforms about 20 classical and state-of-the-art transfer-learning methods in the online setting and achieves accuracy comparable to the best offline methods that see all target data in advance. If this holds, a new user could begin controlling a BCI immediately, with per-trial adaptation replacing the usual calibration block.

What carries the argument

The machinery has three interacting parts. First, incremental Euclidean alignment (IEA) aligns each new target trial using the running mean covariance of all trials seen so far, so the distribution shift between source subjects and the new user is partially removed before classification. Second, the prediction is an ensemble: up to trial M the M classifiers' probability vectors are averaged, and afterwards they are combined by the spectral meta-learner (SML), which forms the sample covariance matrix of the M classifiers' class-wise prediction probabilities over the test stream and uses the principal eigenvector as accuracy-proportional weights. Third, each classifier is updated on a sliding batch of B trials with a loss composed of conditional entropy minimization (with temperature scaling factor T=2) and an adaptive marginal distribution regularizer that estimates the target class-frequency by confidence-thresholded pseudo-labels and renormalizes the batch's mean prediction probabilities, preventing the entropy loss from collapsing to one class. The complete algorithm alternates prediction then update for every arriving trial, so the inference delay is just model forward-pass time while the update runs in the background.

What would settle it

Compare SML-soft weighting against simple averaging of the same M models on the same test streams across all subjects in the three datasets; if averaging wins or ties in most comparisons, the meta-learner's theoretical basis is empirically unsupported.

Watch

Extended reading notes

Core claim

The central claim is that an ensemble of source-trained classifiers, each independently adapted online on the unlabeled target stream, can be combined by a spectral meta-learner—using only the classifiers' own prediction probabilities on test data—to match or beat offline transfer-learning accuracy for cross-subject motor imagery. On the paper's numbers, T-TIME (5) reaches average accuracies of 79.30%, 79.81%, and 78.20% on BNCI2014001, BNCI2014002, and BNCI2015001, respectively, while the best offline approach (MCC) reaches 78.84%, 80.09%, and 78.85%, so the online method lands within about one point of the offline upper reference. The paper also claims robustness under test-time class imbalance, where T-TIME (5) achieves the highest AUC on all three datasets, and reports that continual adaptation across sessions (TTA1+2) gives the best cross-session accuracy. The authors state this is the first study applying test-time adaptation to calibration-free EEG-based BCIs.

Load-bearing premise

The load-bearing premise is that the spectral meta-learner's weighting rule, proven for binary classifiers producing 0/1 predictions under conditional independence, still yields accuracy-proportional weights when applied to multi-class softmax probabilities from correlated deep networks; the paper extends SML to this regime without a proof, so if that extension fails the ensemble may be no better than simple averaging.

Editorial extensions

If this is right

A new BCI user can be served immediately: each trial is classified before adaptation, and the model update runs in parallel with downstream tasks, eliminating the calibration session.
Online TTA accuracy lands within about one point of offline UDA/SFUDA methods that use all target data, so the online constraint does not cost much accuracy in these motor-imagery benchmarks.
Under a 2:1 test-time class imbalance, T-TIME keeps the highest AUC on all three datasets, suggesting the adaptive marginal regularizer protects against skewed target distributions.
Continual cross-session adaptation (adapt to session 1, keep adapting into session 2) gives the best accuracy on the second session, indicating the method handles between-session EEG non-stationarity.

Reading between the lines

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

The unproven multi-class extension of SML is the part I would test first: a controlled comparison of SML-soft against plain averaging on the same model pool could settle whether the meta-learner earns its complexity.
The paper's own future-work list points to natural next experiments: applying T-TIME to event-related-potential or affective BCIs, and to streams without known trial boundaries, where the update would need to segment continuous EEG.
The reported per-model update cost (34–61 ms on a CPU) implies the ensemble size M is bounded by the between-trial interval; a BCI with short inter-trial gaps would have to trade ensemble diversity against update latency.

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.

Desk Editor's Note

read the letter

The central claim holds up: T-TIME is a credible demonstration that test-time adaptation can remove the calibration step in motor-imagery BCIs. The novelty is real — first TTA for EEG-based BCIs, incremental EA, the adaptive marginal-distribution regularizer — and the evaluation is the most careful I've seen in this subfield. Three public datasets, leave-one-subject-out, about 20 baselines, ablations, sensitivity analyses, a continual-TTA extension, and public code. The gains over the strongest online baselines are a few accuracy points but they are consistent across datasets, and the ablation shows each component earns its keep. This deserves a serious referee.

Now the soft spots, in proportion. The biggest is the SML ensemble. Section III-D explicitly extends SML from binary 0/1 predictions to multi-class softmax probabilities without a proof, and the original guarantee from Parisi et al. depends on binary outputs and conditional independence. After T-TIME fine-tunes all M models on the same target batches, those classifiers are correlated, so the rank-one approximation and the eigenvector-to-accuracy link are unsupported. Figure 4, which compares SML-soft to averaging and voting, has no error bars, and Table VI does not isolate SML in the ablations. So the claimed mechanism is not established. That said, this is not fatal: even if SML-soft degrades to a weighted average, the overall method still works, since the rest of the algorithm (CEM + MDR + IEA) carries the performance. The fix is manageable — either prove the extension under weaker assumptions, or explicitly demote SML to a heuristic with empirical support. As written, the paper overclaims when it says SML constructs a meta-classifier more accurate than most.

Second, there are no cross-subject statistical significance tests. The differences between T-TIME and SAR/DELTA are small, and subject-level variance in EEG is large. A paired test or at least confidence intervals on the average would help. This is a moderate issue, not a deal-breaker.

Third, the 'plug-and-play' framing is a bit strong: offline simulations on benchmark data, even with streaming, are not real-time BCI with feedback. The authors do acknowledge this in the computational-cost discussion, so it is minor.

The citation pattern looks fine; self-citing EA and the group's earlier review work is appropriate here. The code and data being public is a genuine strength.

My recommendation: send this out for peer review. It is exactly the kind of empirically grounded, reproducible study that a good referee can make better. I would ask for a substantive revision addressing the SML justification and adding significance tests, then accept.

Referee Report

3 major / 6 minor

Summary. The paper proposes T-TIME, an online test-time adaptation (TTA) method for calibration-free EEG-based BCIs. It trains M EEGNet classifiers on Euclidean-aligned source subjects, then for each incoming target trial performs incremental Euclidean alignment, predicts the label via a spectral meta-learner ensemble, and updates all models using conditional entropy minimization and an adaptive marginal distribution regularizer. The authors evaluate on three public motor imagery datasets under leave-one-subject-out cross-validation, comparing with about 20 UDA/SFUDA/TTA baselines, and report that T-TIME outperforms online baselines and is comparable to the best offline approaches, with additional experiments on class imbalance, ensemble strategies, ablations, parameter sensitivity, continual TTA, and computational cost.

Significance. If the empirical claims hold, T-TIME is a practically useful contribution: it addresses a real gap by moving EEG transfer learning from offline settings to online test-time adaptation, and the evaluation is extensive and careful: three public datasets, leave-one-subject-out, multiple runs, many baselines from different transfer-learning families, sensitivity analyses, and a continual-TTA extension. The public code and the use of MOABB preprocessing support reproducibility. The main caveats are the unproven extension of the Spectral Meta-Learner to multi-class softmax outputs and the absence of statistical significance testing; these do not necessarily undermine the overall contribution if addressed, but they currently leave the core ensemble mechanism and the strongest comparative claims not fully established.

major comments (3)

[Section III-D, Eqs. (5)-(6)] The extension of the Spectral Meta-Learner (SML) to multi-class softmax probabilities is asserted without a supporting argument. The original SML in [37] guarantees that the principal eigenvector of the classifier covariance is proportional to balanced accuracies for binary 0/1 classifiers under a conditional-independence assumption. Here the entries F_k are continuous softmax probabilities from deep networks, and the M models are fine-tuned on the same target batches (Section III-E), so the binary-output and independence conditions are both violated. The note in Section III-F that the models are "independently initialized and updated" does not restore conditional independence because every model is updated with the same unlabeled target batch through the same objective (7)-(13). If the rank-one or proportionality property fails, Eq. (6) has no theoretical basis and may be no better than averaging. Figure 4 shows a comparison to averaging but is reported without error bars or significance tests, and Table VI does not isolate SML. The authors should either provide a proof or a carefully stated heuristic for the multi-class softmax case, or reframe the claim to acknowledge that SML-soft is an empirical heuristic.
[Tables II-V] The paper claims that T-TIME "outperformed about 20 classical and state-of-the-art TL approaches," but no statistical significance test is reported. The tables give mean ± standard deviation over five random restarts, but the unit of the leave-one-subject-out evaluation is the subject; paired tests across subjects (e.g., Wilcoxon signed-rank) are needed to support such comparisons, especially when margins are small. For example, on BNCI2014002 (Table III), T-TIME (5) achieves 79.81% versus 80.09% for the offline MCC, and the single-model T-TIME achieves 78.26% versus 77.53% for SAR. Without a test, the headline claim is not fully substantiated. Please add per-subject paired significance tests and report the number of subjects for which each comparison favors T-TIME.
[Section III-E and Table VI] The contribution of the adaptive marginal distribution regularization over a standard uniform information-maximization term is not isolated. Table VI ablates CEM, MDR, and TR, but MDR always denotes the adaptive version; there is no comparison against the non-adaptive IM loss used in SHOT [20] with uniform target class frequencies. Since the adaptive MDR is one of the paper's stated contributions for class-imbalanced streams, the ablation should include the uniform-IM variant, and the class-imbalance experiments in Table V should compare T-TIME with and without the adaptive reweighting. The pseudo-label count z_k in Eq. (10) is also used as an estimate of the true target class frequency without any robustness analysis; the sensitivity of T-TIME to inaccurate z_k should be discussed or tested (varying the threshold tau in Fig. 5 is only an indirect test).

minor comments (6)

[Algorithm 1] The pseudo-code header contains a typo: "Target Lata Prediction" should be "Target Label Prediction."
[Section IV-B] The value of the sliding batch size B is never stated explicitly in the experimental section; from context it appears to be 8 (the test batch size), but it should be declared explicitly in Section IV-B or in Algorithm 1.
[Section III-D, Eq. (5)] The notation E[F_k(\tilde X_t)] denotes an expectation, but in the algorithm it is computed as the sample mean over the a available trials; please use an explicit average notation for clarity.
[Section III-D, Eq. (6)] Principal eigenvectors are determined only up to sign; the authors should state how the entries of v_k are made non-negative or how the sign ambiguity is resolved before weighting.
[Figure 4] Since the experiments in Fig. 4 were repeated 10 times, please include error bars or shaded confidence bands; the current figure, as described, does not allow the reader to judge the reliability of the SML-soft advantage over Averaging.
[Section IV-C, observation 2] The text says EA "significantly improved" classification performance, but no significance test is reported; please either add tests or soften the wording to "improved."

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: T-TIME's central claim is an external benchmark comparison, and the unproven SML-soft extension is a correctness risk rather than a circular reduction.

full rationale

T-TIME's central claim is an empirical accuracy comparison against about 20 external baselines on three public MOABB datasets, evaluated on held-out target subjects, so no fitted parameter or pseudo-label is renamed as a prediction. The only self-referential components are the SML ensemble in Eqs. (5)-(6), which weights classifiers by the principal eigenvector of the covariance of their own probability outputs, and the pseudo-label frequency z_k in Eq. (10) used in adaptive marginal regularization. Both are standard unsupervised/test-time mechanisms: the model's own outputs are used as soft targets for a loss on unlabeled data, while the reported accuracies are computed against ground-truth labels of held-out target subjects, not against the model's own outputs. The paper's self-citations, particularly Euclidean alignment [12] and the authors' TL reviews [4], [8], supply preprocessing tools and background from prior published work; they do not provide the numerical result, nor do they invoke a uniqueness theorem that forces the outcome. The extension of SML from binary 0/1 predictions to multi-class softmax probabilities in Section III-D is asserted without proof and is a legitimate correctness concern, but it is not circularity: the eigenvector weights are not fitted to target labels, and if the rank-one or accuracy-proportionality claim fails for correlated softmax outputs, the ensemble may underperform simple averaging. The hyperparameter sensitivity analysis is performed on the same datasets, which is a methodological caution, but not a circular derivation. Accordingly, no load-bearing step reduces by construction to its own inputs.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

T-TIME introduces no new physical entities. The method relies on standard closed-set domain adaptation, the borrowed SML ensemble theory (extended without proof), pseudo-label based class frequency estimates, and author-chosen hyperparameters. The key formal gap is the unproven multi-class soft extension of SML.

free parameters (5)

M (number of base classifiers) = 5 (with 1 as an ablation)
Ensemble size; chosen by authors, performance varies with M (Fig. 4).
B (sliding batch size) = 8
Size of the target update batch; chosen by authors.
T (temperature scaling factor) = 2
Recalibrates prediction confidence in Eq. (8) and (9); sensitivity analysis shows T in {2,3,4,5} works.
tau (pseudo-labeling confidence threshold) = 0.7
Threshold in Eq. (10) for estimating target class frequency; sensitivity analysis supports tau in [0.6, 0.8].
c (small integer in adaptive MDR) = 4
Constant in Eq. (11), set to half the batch size; avoids division by zero and scales q_k.

assumptions (4)

domain assumption Source and target domains share identical input and label spaces, differing only in marginal and conditional distributions.
Stated in Section III as the closed-set domain adaptation setting.
ad hoc to paper The SML theoretical guarantee (principal eigenvector weights are proportional to classifier accuracies) extends to multi-class continuous softmax predictions from correlated deep networks.
Section III-D adapts SML to multi-class soft probabilities without proof; original SML [37] assumes binary 0/1 outputs and conditional independence.
ad hoc to paper The pseudo-label based class count z_k in Eq. (10) estimates the true target class frequency well enough for the adaptive MDR to improve adaptation.
No bias or noise analysis is given for the pseudo-label frequency estimate; it depends on the model's own confidence.
domain assumption Euclidean alignment (EA) reduces inter-subject EEG differences sufficiently that a model trained on aligned source data is a useful starting point for a new subject.
Borrowed from prior work [12]; the paper's own experiments show EA is essential for all methods.

how reviews work

0 comments

Cite this review

Pith. "Pith review of T-TIME: Test-Time Information Maximization Ensemble for Plug-and-Play BCIs." pith.science (2026). https://pith.science/paper/73DQ4MMY

@misc{pith2026241207228,
  author       = {Pith},
  title        = {Pith review of: T-TIME: Test-Time Information Maximization Ensemble for Plug-and-Play BCIs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/73DQ4MMY}},
  note         = {Machine review of arXiv:2412.07228}
}

read the original abstract

Objective: An electroencephalogram (EEG)-based brain-computer interface (BCI) enables direct communication between the human brain and a computer. Due to individual differences and non-stationarity of EEG signals, such BCIs usually require a subject-specific calibration session before each use, which is time-consuming and user-unfriendly. Transfer learning (TL) has been proposed to shorten or eliminate this calibration, but existing TL approaches mainly consider offline settings, where all unlabeled EEG trials from the new user are available. Methods: This paper proposes Test-Time Information Maximization Ensemble (T-TIME) to accommodate the most challenging online TL scenario, where unlabeled EEG data from the new user arrive in a stream, and immediate classification is performed. T-TIME initializes multiple classifiers from the aligned source data. When an unlabeled test EEG trial arrives, T-TIME first predicts its labels using ensemble learning, and then updates each classifier by conditional entropy minimization and adaptive marginal distribution regularization. Our code is publicized. Results: Extensive experiments on three public motor imagery based BCI datasets demonstrated that T-TIME outperformed about 20 classical and state-of-the-art TL approaches. Significance: To our knowledge, this is the first work on test time adaptation for calibration-free EEG-based BCIs, making plug-and-play BCIs possible.

Figures

Figures reproduced from arXiv: 2412.07228 by the authors.

**Figure 2.** Flowchart of the proposed T-TIME. B. Source Model Training We perform Euclidean alignment (EA) [12] on each source subject individually to reduce the individual differences, and then combine all aligned EEG trials from the L source subjects into a single source domain to train M source models. For the l-th (l = 1, ..., L) source subject, EA computes the arithmetic mean of all covariance matrices of his/her EEG trial… view at source ↗

**Figure 3.** Conditional entropy minimization and adaptive marg [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: With different random seeds and training batch split [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: Performance of T-TIME w.r.t. different hyperparame [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

**Figure 6.** Figure 6: Two portions of computation time of T-TIME. The compu [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 36 canonical work pages

[37]

Ranking and combining multiple predictors without labeled data,

F. Parisi, F. Strino, B. Nadler, and Y . Kluger, “Ranking and combining multiple predictors without labeled data,” Proc. National Academy of Sciences, vol. 111, no. 4, pp. 1253–1258, 2014

work page 2014
[20]

Source data- absent unsupervised domain adaptation through hypothesis transf er and label- ing transfer,

J. Liang, D. Hu, Y . Wang, R. He, and J. Feng, “Source data- absent unsupervised domain adaptation through hypothesis transf er and label- ing transfer,” IEEE Trans. Pattern Analysis and Machine Intelligence , vol. 44, no. 11, pp. 8602–8617, 2022

work page 2022
[1]

Brain-c omputer in- terfaces: A gentle introduction,

B. Graimann, B. Allison, and G. Pfurtscheller, “Brain-c omputer in- terfaces: A gentle introduction,” Brain-Computer Interfaces , pp. 1–27, 2009

work page 2009
[2]

Enhancing nervous system recovery through neurob iologics, neural interface training, and neurorehabilitation,

M. O. Krucoff, S. Rahimpour, M. W. Slutzky, V . R. Edgerton , and D. A. Turner, “Enhancing nervous system recovery through neurob iologics, neural interface training, and neurorehabilitation,” Frontiers in Neuro- science, vol. 10, p. 584, 2016

work page 2016
[3]

Brain-computer interfaces: Beyond medical applications,

J. van Erp, F. Lotte, and M. Tangermann, “Brain-computer interfaces: Beyond medical applications,” Computer, vol. 45, no. 4, pp. 26–34, 2012

work page 2012
[4]

Transfer learning for EEG-bas ed brain–computer interfaces: A review of progress made since 2016,

D. Wu, Y . Xu, and B.-L. Lu, “Transfer learning for EEG-bas ed brain–computer interfaces: A review of progress made since 2016,” IEEE Trans. Cognitive and Developmental Systems , vol. 14, no. 1, pp. 4–19, 2022

work page 2016
[5]

Evoked-pote ntial correlates of stimulus uncertainty,

S. Sutton, M. Braren, J. Zubin, and E. John, “Evoked-pote ntial correlates of stimulus uncertainty,” Science, vol. 150, no. 3700, pp. 1187–1188, 1965

work page 1965
[6]

Multiple channel detection of steady-state visual evoked potentials for brain-comput er interfaces,

O. Friman, I. V olosyak, and A. Graser, “Multiple channel detection of steady-state visual evoked potentials for brain-comput er interfaces,” IEEE Trans. Biomedical Engineering , vol. 54, no. 4, pp. 742–750, 2007

work page 2007

Show all 44 references

[7]

Motor imagery and direc t brain- computer communication,

G. Pfurtscheller and C. Neuper, “Motor imagery and direc t brain- computer communication,” Proc. IEEE , vol. 89, no. 7, pp. 1123–1134, 2001

2001
[8]

Transfer learning for motor imagery based brain-computer interfaces: A tutorial,

D. Wu, X. Jiang, and R. Peng, “Transfer learning for motor imagery based brain-computer interfaces: A tutorial,” Neural Networks, vol. 153, pp. 235–253, 2022

2022
[9]

A survey of unsupervised deep do main adaptation,

G. Wilson and D. J. Cook, “A survey of unsupervised deep do main adaptation,” ACM Trans. Intelligent Systems and Technology , vol. 11, no. 5, pp. 1–46, 2020

2020
[10]

Return of frustratingly easy domain adaptation,

B. Sun, J. Feng, and K. Saenko, “Return of frustratingly easy domain adaptation,” in Proc. AAAI Conf. Artiﬁcial Intelligence , Phoenix, AZ, Feb. 2016, pp. 2058–2065

2016
[11]

Trans- fer learning: A Riemannian geometry framework with applica tions to brain–computer interfaces,

P . Zanini, M. Congedo, C. Jutten, S. Said, and Y . Berthou mieu, “Trans- fer learning: A Riemannian geometry framework with applica tions to brain–computer interfaces,” IEEE Trans. Biomedical Engineering , vol. 65, no. 5, pp. 1107–1116, 2018

2018
[12]

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

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

2020
[13]

Learning tran sferable features with deep adaptation networks,

M. Long, Y . Cao, J. Wang, and M. I. Jordan, “Learning tran sferable features with deep adaptation networks,” in Proc. Int’l Conf. Machine Learning, Lille, France, Jul. 2015, pp. 97–105

2015
[14]

Deep transfer learning with joint adaptation networks,

M. Long, H. Zhu, J. Wang, and M. I. Jordan, “Deep transfer learning with joint adaptation networks,” in Proc. Int’l Conf. Machine Learning , Sydney, Australia, Aug. 2017, pp. 2208–2217

2017
[15]

Domain-adversarial training of neural networks,

Y . Ganin, E. Ustinova, H. Ajakan, P . Germain, H. Laroche lle, F. Lavio- lette, M. Marchand, and V . Lempitsky, “Domain-adversarial training of neural networks,” Journal of Machine Learning Research , vol. 17, no. 1, pp. 2096–2030, 2016

2016
[16]

Conditional a dversarial domain adaptation,

M. Long, Z. Cao, J. Wang, and M. I. Jordan, “Conditional a dversarial domain adaptation,” in Proc. Advances Neural Information Processing Systems, Montreal, Canada, Dec. 2018, pp. 1640–1650

2018
[17]

Minimum class conf usion for versatile domain adaptation,

Y . Jin, X. Wang, M. Long, and J. Wang, “Minimum class conf usion for versatile domain adaptation,” in Proc. European Conf. Computer Vision , Glasgow, UK, Aug. 2020, pp. 464–480

2020
[18]

Domain adaptation with aux iliary target domain-oriented classiﬁer,

J. Liang, D. Hu, and J. Feng, “Domain adaptation with aux iliary target domain-oriented classiﬁer,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition , Nashville, TN, Jun. 2021, pp. 16 632–16 642

2021
[19]

Source-fr ee unsuper- vised domain adaptation: A survey,

Y . Fang, P .-T. Y ap, W. Lin, H. Zhu, and M. Liu, “Source-fr ee unsuper- vised domain adaptation: A survey,” arXiv preprint arXiv:2301.00265 , 2022

2022 arXiv
[21]

Imbalanced so urce-free domain adaptation,

X. Li, J. Li, L. Zhu, G. Wang, and Z. Huang, “Imbalanced so urce-free domain adaptation,” in Proc. ACM Int’l Conf. Multimedia , Chengdu, China, Oct. 2021, pp. 3330–3339

2021
[22]

Privacy-preserving brain–com puter interfaces: A systematic review,

K. Xia, W. Duch, Y . Sun, K. Xu, W. Fang, H. Luo, Y . Zhang, D. Sang, X. Xu, F.-Y . Wang, and D. Wu, “Privacy-preserving brain–com puter interfaces: A systematic review,” IEEE Trans. Computational Social Systems, early access, 2022

2022
[23]

Meta-learning for fast a nd privacy-preserving source knowledge transfer of EEG-base d BCIs,

S. Li, H. Wu, L. Ding, and D. Wu, “Meta-learning for fast a nd privacy-preserving source knowledge transfer of EEG-base d BCIs,” IEEE Computational Intelligence Magazine , vol. 17, no. 4, pp. 16–26, 2022

2022
[24]

Lightweight source-free transfer f or privacy- preserving motor imagery classiﬁcation,

W. Zhang and D. Wu, “Lightweight source-free transfer f or privacy- preserving motor imagery classiﬁcation,” IEEE Trans. Cognitive and Developmental Systems , early access, 2022

2022
[25]

Multi-source decentraliz ed transfer for privacy-preserving BCIs,

W. Zhang, Z. Wang, and D. Wu, “Multi-source decentraliz ed transfer for privacy-preserving BCIs,” IEEE Trans. Neural Systems and Reha- bilitation Engineering , vol. 30, pp. 2710–2720, 2022

2022
[26]

A comprehensive survey on te st-time adaptation under distribution shifts,

J. Liang, R. He, and T. Tan, “A comprehensive survey on te st-time adaptation under distribution shifts,” arXiv preprint arXiv:2303.15361 , 2023

2023 arXiv
[27]

Pseudo-label: The simple and efﬁcient semi -supervised learning method for deep neural networks,

D.-H. Lee, “Pseudo-label: The simple and efﬁcient semi -supervised learning method for deep neural networks,” in Proc. Int’l Conf. Machine Learning W orkshops, Atlanta, GA, Jun. 2013, pp. 1322–1333

2013
[28]

Contrast ive test-time adaptation,

D. Chen, D. Wang, T. Darrell, and S. Ebrahimi, “Contrast ive test-time adaptation,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition, New Orleans, LA, Jun. 2022, pp. 295–305

2022
[29]

Test-time classiﬁer adjustm ent module for model-agnostic domain generalization,

Y . Iwasawa and Y . Matsuo, “Test-time classiﬁer adjustm ent module for model-agnostic domain generalization,” in Proc. Advances Neural Information Processing Systems , Virtual, Dec. 2021, pp. 2427–2440

2021
[30]

Tent: Fully test-time adaptation by entropy minimization,

D. Wang, E. Shelhamer, S. Liu, B. Olshausen, and T. Darre ll, “Tent: Fully test-time adaptation by entropy minimization,” in Proc. Int’l Conf. Learning Representations, Vienna, Austria, May. 2021

2021
[31]

MEMO: Test time robu stness via adaptation and augmentation,

M. M. Zhang, S. Levine, and C. Finn, “MEMO: Test time robu stness via adaptation and augmentation,” in Proc. Advances Neural Information Processing Systems, New Orleans, LA, Nov. 2022, pp. 38 629–38 642

2022
[32]

Towards stable test-time adaptation in dynamic wild world,

S. Niu, J. Wu, Y . Zhang, Z. Wen, Y . Chen, P . Zhao, and M. Tan, “Towards stable test-time adaptation in dynamic wild world,” in Proc. Int’l Conf. Learning Representations, Kigali, Rwanda, May. 2023

2023
[33]

DELTA: degradation-fr ee fully test- time adaptation,

B. Zhao, C. Chen, and S.-T. Xia, “DELTA: degradation-fr ee fully test- time adaptation,” in Proc. Int’l Conf. Learning Representations , Kigali, Rwanda, May. 2023

2023
[34]

Continual test -time domain adaptation,

Q. Wang, O. Fink, L. V an Gool, and D. Dai, “Continual test -time domain adaptation,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition, New Orleans, LA, Jun. 2022, pp. 7201–7211

2022
[35]

Continual prototype evo lution: Learn- ing online from non-stationary data streams,

M. De Lange and T. Tuytelaars, “Continual prototype evo lution: Learn- ing online from non-stationary data streams,” in Proc. IEEE/CVF Int’l Conf. Computer Vision , Montreal, Canada, Oct. 2021, pp. 8250–8259

2021
[36]

EEGNet: A compact convolutional neural net work 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 net work for EEG-based brain-computer interfaces,” Journal of Neural Engineering , vol. 15, no. 5, p. 056013, 2018

2018
[38]

On calib ration of modern neural networks,

C. Guo, G. Pleiss, Y . Sun, and K. Q. Weinberger, “On calib ration of modern neural networks,” in Proc. Int’l Conf. Machine Learning , Sydney, Australia, Aug. 2017, pp. 1321–1330

2017
[39]

Parameter-free online test-time adaptation,

M. Boudiaf, R. Mueller, I. Ben Ayed, and L. Bertinetto, “ Parameter-free online test-time adaptation,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition , New Orleans, LA, Jun. 2022, pp. 8344–8353

2022
[40]

MOABB: trustworthy algor ithm bench- marking for BCIs,

V . Jayaram and A. Barachant, “MOABB: trustworthy algor ithm bench- marking for BCIs,” Journal of Neural Engineering , vol. 15, no. 6, p. 066011, 2018

2018
[41]

Optimizing spatial ﬁlters for robust EEG single-trial ana lysis,

B. Blankertz, R. Tomioka, S. Lemm, M. Kawanabe, and K.-r . Muller, “Optimizing spatial ﬁlters for robust EEG single-trial ana lysis,” IEEE Signal Processing Magazine , vol. 25, no. 1, pp. 41–56, 2008

2008
[42]

Bridging theor y and algorithm for domain adaptation,

Y . Zhang, T. Liu, M. Long, and M. Jordan, “Bridging theor y and algorithm for domain adaptation,” in Proc. Int’l Conf. Machine Learning , Long Beach, CA, Jun. 2019, pp. 7404–7413. 11

2019
[43]

Improving robustness against common corruptio ns by covariate shift adaptation,

S. Schneider, E. Rusak, L. Eck, O. Bringmann, W. Brendel , and M. Bethge, “Improving robustness against common corruptio ns by covariate shift adaptation,” in Proc. Advances Neural Information Pro- cessing Systems , V ancouver, Canada, Dec. 2020, pp. 11 539–11 551

2020
[44]

Dynamically weighted ensemble classiﬁcation for non-sta tionary EEG processing,

S. R. Liyanage, C. Guan, H. Zhang, K. K. Ang, J. Xu, and T. H . Lee, “Dynamically weighted ensemble classiﬁcation for non-sta tionary EEG processing,” Journal of Neural Engineering , vol. 10, no. 3, p. 036007, 2013

2013

Pith tools

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

[1] [37]

Ranking and combining multiple predictors without labeled data,

F. Parisi, F. Strino, B. Nadler, and Y . Kluger, “Ranking and combining multiple predictors without labeled data,” Proc. National Academy of Sciences, vol. 111, no. 4, pp. 1253–1258, 2014

work page 2014

[2] [20]

Source data- absent unsupervised domain adaptation through hypothesis transf er and label- ing transfer,

J. Liang, D. Hu, Y . Wang, R. He, and J. Feng, “Source data- absent unsupervised domain adaptation through hypothesis transf er and label- ing transfer,” IEEE Trans. Pattern Analysis and Machine Intelligence , vol. 44, no. 11, pp. 8602–8617, 2022

work page 2022

[3] [1]

Brain-c omputer in- terfaces: A gentle introduction,

B. Graimann, B. Allison, and G. Pfurtscheller, “Brain-c omputer in- terfaces: A gentle introduction,” Brain-Computer Interfaces , pp. 1–27, 2009

work page 2009

[4] [2]

Enhancing nervous system recovery through neurob iologics, neural interface training, and neurorehabilitation,

M. O. Krucoff, S. Rahimpour, M. W. Slutzky, V . R. Edgerton , and D. A. Turner, “Enhancing nervous system recovery through neurob iologics, neural interface training, and neurorehabilitation,” Frontiers in Neuro- science, vol. 10, p. 584, 2016

work page 2016

[5] [3]

Brain-computer interfaces: Beyond medical applications,

J. van Erp, F. Lotte, and M. Tangermann, “Brain-computer interfaces: Beyond medical applications,” Computer, vol. 45, no. 4, pp. 26–34, 2012

work page 2012

[6] [4]

Transfer learning for EEG-bas ed brain–computer interfaces: A review of progress made since 2016,

D. Wu, Y . Xu, and B.-L. Lu, “Transfer learning for EEG-bas ed brain–computer interfaces: A review of progress made since 2016,” IEEE Trans. Cognitive and Developmental Systems , vol. 14, no. 1, pp. 4–19, 2022

work page 2016

[7] [5]

Evoked-pote ntial correlates of stimulus uncertainty,

S. Sutton, M. Braren, J. Zubin, and E. John, “Evoked-pote ntial correlates of stimulus uncertainty,” Science, vol. 150, no. 3700, pp. 1187–1188, 1965

work page 1965

[8] [6]

Multiple channel detection of steady-state visual evoked potentials for brain-comput er interfaces,

O. Friman, I. V olosyak, and A. Graser, “Multiple channel detection of steady-state visual evoked potentials for brain-comput er interfaces,” IEEE Trans. Biomedical Engineering , vol. 54, no. 4, pp. 742–750, 2007

work page 2007