Transfer Learning for Brain–Computer Interfaces: A Euclidean Space Data Alignment Approach
Open accessThe DOI matches its Crossref record; 50 of the 52 references it cites were found in a registry or an index; 9 of the 30 sources its claims cite were checked for support.
Abstract
OBJECTIVE: This paper targets a major challenge in developing practical electroencephalogram (EEG)-based brain-computer interfaces (BCIs): how to cope with individual differences so that better learning performance can be obtained for a new subject, with minimum or even no subject-specific data? METHODS: We propose a novel approach to align EEG trials from different subjects in the Euclidean space to make them more similar, and hence improve the learning performance for a new subject. Our approach has three desirable properties: first, it aligns the EEG trials directly in the Euclidean space, and any signal processing, feature extraction, and machine learning algorithms can then be applied to the aligned trials; second, its computational cost is very low; and third, it is unsupervised and does not need any label information from the new subject. RESULTS: Both offline and simulated online experiments on motor imagery classification and event-related potential classification verified that our proposed approach outperformed a state-of-the-art Riemannian space data alignment approach, and several approaches without data alignment. CONCLUSION: The proposed Euclidean space EEG data alignment approach can greatly facilitate transfer learning in BCIs. SIGNIFICANCE: Our proposed approach is effective, efficient, and easy to implement. It could be an essential pre-processing step for EEG-based BCIs.
Overview
CoChat Summary
TL;DR
The paper proposes Euclidean-space alignment (EA), an unsupervised transfer-learning preprocessing step that aligns EEG trials from different subjects in Euclidean space to reduce individual differences for BCI classification. Across offline and simulated online experiments on motor imagery and ERP datasets, EA generally outperformed Riemannian-space alignment and no-alignment baselines while being much faster.
Key findings
- In offline unsupervised MI classification, EA-CSP-LDA achieved average accuracies of 79.79% on MI Dataset 1 and 73.53% on MI Dataset 2a, versus RA-MDRM at 68.07% and 70.91%, CSP-LDA at 59.71% and 67.75%, and MDRM at 54.36% and 57.18%.
- In offline ERP classification, EA-xDAWN-SVM achieved an average balanced classification accuracy (BCA) of 68.80%, compared with EA-SVM 67.85%, SVM 64.64%, and xDAWN-SVM 64.60%; EA-SVM significantly outperformed SVM (p=0.0386) and EA-xDAWN-SVM significantly outperformed xDAWN-SVM (p=0.0449).
- In simulated online MI classification, EA-CSP-LDA outperformed RA-MDRM on 12 out of 16 subjects; in simulated online ERP classification, EA-xDAWN-SVM outperformed RA-MDRM on all 11 subjects.
- EA-CSP-LDA was 3.6–19.5 times faster than RA-MDRM; for example, on MI Dataset 1 it took 0.3864 s versus 7.5326 s, and on MI Dataset 2a it took 0.2405 s versus 0.8766 s.
Methods
- The proposed EA aligns time-domain EEG trials in Euclidean space by using the arithmetic mean of all covariance matrices from a subject as the reference matrix, then transforming trials so that the mean covariance matrix of each subject becomes the identity matrix; it requires no labels from the new subject.
- Experiments used two MI datasets from BCI Competition IV (Dataset 1: 7 subjects, 59-channel EEG, 100 trials per class in calibration; Dataset 2a: 9 subjects, 22-channel EEG plus 3-channel EOG, 72 trials per class for left- and right-hand MI) and one ERP RSVP dataset from PhysioNet (11 subjects, 8 channels, 5 Hz Session a, target-to-non-target ratio around 1:9).
- Evaluation included offline unsupervised leave-one-subject-out cross-validation and simulated online supervised classification, with CSP-LDA, MDRM, SVM, and xDAWN-SVM pipelines; paired-sample t-tests and area under the curve (AUC) were used for statistical comparisons.
Limitations
- The paper states that EA only compensates for covariate shift and ignores prior probability shift and concept shift; it may even increase concept shift, and bad trials or outliers can corrupt the reference matrix. It also states that the simulated online experiments are not identical to real online experiments.
- The text makes evident that EA degraded accuracy for some subjects (e.g., Subjects 4 and 9 in MI Dataset 2a, and some ERP subjects), and that RA-MDRM could not be applied to offline unsupervised ERP classification because it requires labeled non-target and target trials from the new subject.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
4 Valid · 8 Not valid · 15 Not checked · 89 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“So Barachant and Congedo [2] proposed a novel approach to augment the ERP trials so that their covariance matrices can also encode the temporal information.”
What the sources say
- Supportedcites [2] [2] A. Barachant and M. Congedo, 'A plug & play P300 BCI using information geometry,' arXiv: 1409.0107 , 2014. - source
Section B (pp. 3–4) describes a prototyped ERP response obtained by averaging single-trial responses, then concatenating it with each trial to form a super trial. The resulting covariance matrices include a cross-covariance block that carries information about the ERP waveform’s temporal structure. (LLM, conf=0.94)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“However, the main discriminative information of ERP trials is carried temporally rather than spatially.”
What the sources say
- Supportedcites [2] [2] A. Barachant and M. Congedo, 'A plug & play P300 BCI using information geometry,' arXiv: 1409.0107 , 2014. - source
Section B (p. 3) says ERP discriminative information lies in the potential’s waveform rather than the spatial distribution of its variance. It also says ignoring temporal structure loses information needed for classification, supporting the paraphrase. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our proposed EA-CSP-LDA was 3.6-19.5 times faster than RA-MDRM, and also it had much smaller standard deviation.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“On average RI-MDRM outperformed RR-MDRM, and EI-CSP-LDA outperformed ER-CSP-LDA, on both datasets, suggesting that estimating the reference matrix from all imagery trials would be better than using all resting trials.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our proposed EA-CSP-LDA was 3.6-19.5 times faster than RA-MDRM, and also it had much smaller standard deviation.”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
47 more claims
Create a free account to see all 116 claims and to verify the 4 references still open.
- It may state two separate claims as one. 47
- Probably not a substantial claim. 30
- The quote only partly states this claim. 1
Citations
Cites
LIBSVM: A library for support vector machines
Chang, C.-C., & Lin, C.-J. (2011) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/1961189.1961199 · Cited by 40,918 · cited 1×
“LibSVM [6] with a linear kernel was used for classification.”
Method(2008) · Journal of Machine Learning Research · Cited by 33,051
Delorme, A., & Makeig, S. (2004) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2003.10.009 · Cited by 24,371
Pan, S. J., & Yang, Q. (2010) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2009.191 · Cited by 23,387 · cited 2×
“This idea has been widely used in TL [24], [31], [40].”
BackgroundMethodBook Review: Pattern Recognition and Machine Learning
Nasrabadi, N. M. (2007) · Journal of Electronic Imaging · doi:10.1117/1.2819119 · Cited by 22,066
A Simple Sequentially Rejective Multiple Test Procedure
(1979) · Scandinavian Journal of Statistics · doi:10.2307/4615733 · Cited by 21,526 · cited 1×
“Hence, we performed Holm-Bonferroni corrections [12] on the three p-values of each dataset.”
MethodGoldberger, A. L., Amaral, L. A. N., Glass, L., Hausdorff, J. M., Ivanov, P. C., Mark, R. G., Mietus, J. E., Moody, G. B., Peng, C.-K., & Stanley, H. E. (2000) · Circulation · doi:10.1161/01.cir.101.23.e215 · Cited by 14,809
Pattern Recognition and Machine Learning
(2006) · doi:10.1007/978-0-387-45528-0 · Cited by 9,824 · cited 1×
“Usually the EEG signals are first band-pass filtered and spatially filtered to increase the signal-to-noise ratio, and then discriminative features are extracted, which are next fed into machine learning algorithms such as Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) [3] for classification.”
MethodBrain–computer interfaces for communication and control
Wolpaw, J. R., Birbaumer, N., McFarland, D. J., Pfurtscheller, G., & Vaughan, T. M. (2002) · Clinical Neurophysiology · doi:10.1016/s1388-2457(02)00057-3 · Cited by 7,927
On the Kolmogorov-Smirnov Test for Normality with Mean and Variance Unknown
Lilliefors, H. W. (1967) · Journal of the American Statistical Association · doi:10.2307/2283970 · Cited by 4,530
Cited by
Artificial intelligence for the metaverse: A survey
Huynh-The, T., Pham, Q.-V., Pham, X.-Q., Nguyen, T. T., Han, Z., & Kim, D.-S. (2023) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2022.105581 · Cited by 710
“With electroencephalogram (EEG) signal as one of the most popular inputs of BCI systems, the work in [128] studied a”
MethodA review on transfer learning in EEG signal analysis
Wan, Z., Yang, R., Huang, M., Zeng, N., & Liu, X. (2021) · Neurocomputing · doi:10.1016/j.neucom.2020.09.017 · Cited by 468
Zhang, W., Deng, L., Zhang, L., & Wu, D. (2023) · IEEE/CAA Journal of Automatica Sinica · doi:10.1109/jas.2022.106004 · Cited by 427
Self-citationTransfer Learning for EEG-Based Brain–Computer Interfaces: A Review of Progress Made Since 2016
Wu, D., Xu, Y., & Lu, B.-L. (2022) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2020.3007453 · Cited by 434
“For example, EA and LA data alignment schemes have been combined with TL classifiers [50], [60], and CSDF-ReliefF and CSDF-mRMR feature selection approaches have also been integrated with TL classifiers [114].”
Self-citationProgress in Brain Computer Interface: Challenges and Opportunities
Saha, S., Mamun, K. A., Ahmed, K., Mostafa, R., Naik, G. R., Darvishi, S., Khandoker, A. H., & Baumert, M. (2021) · Frontiers in Systems Neuroscience · doi:10.3389/fnsys.2021.578875 · Cited by 365
Intra- and Inter-subject Variability in EEG-Based Sensorimotor Brain Computer Interface: A Review
Saha, S., & Baumert, M. (2020) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2019.00087 · Cited by 325
“Many works on transfer learning for SMR-based BCI proposed the use of a very few training samples from the target subject (Kang et al., 2009; Lu et al., 2010; Kang and Choi, 2014; Fahimi et al., 2018; He and Wu, 2019).”
BackgroundDeep Representation-Based Domain Adaptation for Nonstationary EEG Classification
Zhao, H., Zheng, Q., Ma, K., Li, H., & Zheng, Y. (2021) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2020.3010780 · Cited by 253
“He and Wu [39] proposed a data alignment framework in the Euclidean space, where the covariance matrices were aligned between different subjects and could be used as features for classification.”
BackgroundCognitive Workload Recognition Using EEG Signals and Machine Learning: A Review
Zhou, Y., Huang, S., Xu, Z., Wang, P., Wu, X., & Zhang, D. (2022) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2021.3090217 · Cited by 221
“…we can also learn lessons from other EEG-based classification tasks, such as motor imagery that adopting the data alignment method as a procedure [150], which aligns EEG trials from different subjects in the Euclidean space to make them more similar; and emotion recognition that using transfer…”
MethodHou, Y., Jia, S., Lun, X., Hao, Z., Shi, Y., Li, Y., Zeng, R., & Lv, J. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2022.3202569 · Cited by 196
Singh, A., Hussain, A. A., Lal, S., & Guesgen, H. W. (2021) · Sensors · doi:10.3390/s21062173 · Cited by 192
“The former can be seen in He and Wu [135], Hossain et al. [136], and Dai et al. [137].”
BackgroundManifold Embedded Knowledge Transfer for Brain-Computer Interfaces
Zhang, W., & Wu, D. (2020) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2020.2985996 · Cited by 183
“Euclidean Alignment (EA) Although RA-MDM has demonstrated promising performance, it still has some limitations [23]: 1) it processes the covariance matrices in the Riemannian space, whereas there are very few Riemannian space classifiers; 2) it computes the reference matrix from the non-target stimuli in ERP-based BCI…”
BackgroundMethodInfluentialSelf-citationMiao, Z., Zhao, M., Zhang, X., & Ming, D. (2023) · NeuroImage · doi:10.1016/j.neuroimage.2023.120209 · Cited by 186
“(2022)), which included trial normalization and Euclidean alignment (He and Wu (2019)).”
Li, Y., Guo, L., Liu, Y., Liu, J., & Meng, F. (2021) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2021.3099908 · Cited by 171
Transfer learning for motor imagery based brain–computer interfaces: A tutorial
Wu, D., Jiang, X., & Peng, R. (2022) · Neural Networks · doi:10.1016/j.neunet.2022.06.008 · Cited by 174
“Different EEG trial alignment approaches have been proposed recently [22], [23], [25], [28], [29].”
BackgroundMethodInfluentialSelf-citationDynamic Joint Domain Adaptation Network for Motor Imagery Classification
Hong, X., Zheng, Q., Liu, L., Chen, P., Ma, K., Gao, Z., & Zheng, Y. (2021) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2021.3059166 · Cited by 140
“For instance, He and Wu [23] aligned EEG trials from different subjects before classification under the assumption that the mean Euclidean covariance matrix should be an identity matrix for all subjects after alignment.”
BackgroundAffective Brain–Computer Interfaces (aBCIs): A Tutorial
Wu, D., Lu, B., Hu, B., & Zeng, Z. (2023) · Proceedings of the IEEE · doi:10.1109/jproc.2023.3277471 · Cited by 151
“EEG data alignment approaches, e.g., Riemannian alignment [125] and Euclidean alignment [126], may also be used to transform the EEG data from different subjects so that their distributions are more consistent (and hence to facilitate transfer learning); thus, they may also be viewed as feature…”
MethodSelf-citationCross-Dataset Variability Problem in EEG Decoding With Deep Learning
Xu, L., Xu, M., Ke, Y., An, X., Liu, S., & Ming, D. (2020) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2020.00103 · Cited by 173
“Instead of direct operation on covariances, we transformed E i + 1 to aligned time series ˆ E i + 1 by The above transformation has also been mentioned in He and Wu (2020), where the authors used Euclidean mean covariance instead of Riemannian mean covariance M here.”
MethodShe, Q., Chen, T., Fang, F. Z., Zhang, J., Gao, Y., & Zhang, Y. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3241846 · Cited by 145
“He and Wu [21] explored a Euclidean alignment (EA) approach which can transform and align MI-EEG signals of different trials in the Euclidean space.”
BackgroundMethodDiscriminative Joint Probability Maximum Mean Discrepancy (DJP-MMD) for Domain Adaptation
Zhang, W., & Wu, D. (2020) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn48605.2020.9207365 · Cited by 127
“Domain adaptation (DA), or transfer learning, has shown promising performance in handling these challenges [2]–[8], by transferring knowledge from a labeled source domain to a new unlabeled or partially labeled target domain.”
Self-citationReview of brain encoding and decoding mechanisms for EEG-based brain–computer interface
Xu, L., Xu, M., Jung, T., & Ming, D. (2021) · Cognitive Neurodynamics · doi:10.1007/s11571-021-09676-z · Cited by 110
“To reduce its computational cost, He and Wu (2019) suggested using the Euclidean mean covariance matrix instead of the Riemannian mean covariance matrix.”
MethodThinker invariance: enabling deep neural networks for BCI across more people
Kostas, D., & Rudzicz, F. (2020) · Journal of Neural Engineering · doi:10.1088/1741-2552/abb7a7 · Cited by 113
“These data consist of 11 subjects presented images at three different rates 5, 6 and 10 Hz, of which we 7 https://www.kaggle.com/c/inria-bci-challenge. use the 5z to match He et al [10].”
BackgroundMethodResultInfluentialToward Open-World Electroencephalogram Decoding Via Deep Learning: A comprehensive survey
Chen, X., Li, C., Liu, A., McKeown, M. J., Qian, R., & Wang, Z. J. (2022) · IEEE Signal Processing Magazine · doi:10.1109/msp.2021.3134629 · Cited by 100
“Statistical feature alignment aims to map the EEG data into a subspace to align the statistical features in the source and target domains [34], such as variance and median absolute deviation.”
BackgroundMI-CAT: A transformer-based domain adaptation network for motor imagery classification
Zhang, D., Li, H., & Xie, J. (2023) · Neural Networks · doi:10.1016/j.neunet.2023.06.005 · Cited by 93
Different Set Domain Adaptation for Brain-Computer Interfaces: A Label Alignment Approach
He, H., & Wu, D. (2020) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2020.2980299 · Cited by 98
“Although it has been shown to significantly improve the cross-subject classification performance in [27], it only aligns the data in the feature space, and may not work well when the source subjects and the target subject have different label spaces.”
BackgroundMethodSelf-citationA Transfer Learning Model for Gesture Recognition Based on the Deep Features Extracted by CNN
Zou, Y., & Cheng, L. (2021) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2021.3098253 · Cited by 96
EEG-Based Driver Drowsiness Estimation Using Feature Weighted Episodic Training
Cui, Y., Xu, Y., & Wu, D. (2019) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2019.2945794 · Cited by 86
“He and Wu [32] proposed a similar EEG data alignment approach in the Euclidean space, which is more efficient than the Riemannian space data alignment approach, and can be used as a pre-processing step before any Euclidean space classifier.”
MethodSelf-citationPrivacy-Preserving Brain–Computer Interfaces: A Systematic Review
Xia, K., Duch, W., Sun, Y., Xu, K., Fang, W., Luo, H., Zhang, Y., Sang, D., Xu, X., Wang, F., et al. (2022) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2022.3184818 · Cited by 114
“One promising solution to reducing the variations is transfer learning [47], [103], [104], which utilizes auxiliary data or knowledge from previous subjects/tasks ( source domains ) to facilitate the calibration for a new subject/task ( target domain ).”
Self-citationA review of artificial intelligence for EEG‐based brain−computer interfaces and applications
Cao, Z. (2020) · Brain Science Advances · doi:10.26599/bsa.2020.9050017 · Cited by 111
Privacy-Preserving Domain Adaptation for Motor Imagery-Based Brain-Computer Interfaces
Xia, K., Deng, L., Duch, W., & Wu, D. (2022) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2022.3168570 · Cited by 79
“Inspired by RA, He and Wu [39] proposed a Euclidean alignment (EA) approach to align the raw EEG trials in the Euclidean space, instead of the covariance matrices in the Riemannian space.”
MethodSelf-citationTIE-EEGNet: Temporal Information Enhanced EEGNet for Seizure Subtype Classification
Peng, R., Zhao, C., Jiang, J., Kuang, G., Cui, Y., Xu, Y., Du, H., Shao, J., & Wu, D. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3204540 · Cited by 66
Self-citationZhang, X., Wu, D., Ding, L., Luo, H., Lin, C.-T., Jung, T., & Chavarriaga, R. (2020) · National Science Review · doi:10.1093/nsr/nwaa233 · Cited by 70
Self-citationGlobal Adaptive Transformer for Cross-Subject Enhanced EEG Classification
Song, Y., Zheng, Q., Wang, Q., Gao, X., & Heng, P. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3285309 · Cited by 68
“Another important issue is that this paper explores domain adaptation for EEG classification from a deep learning perspective, while ignoring the comparison with some traditional methods [41], [42].”
BackgroundZhang, X. L., Wu, D., Ding, L., Luo, H., Lin, C.-T., Jung, T., & Chavarriaga, R. (2020) · Zürcher Hochschule für Angewandte Wissenschaften digital collection (Zurich University of Applied Sciences) · doi:10.21256/zhaw-20507 · Cited by 57
Self-citationChang, H., Zong, Y., Zheng, W., Tang, C., Zhu, J. P., & Li, X. (2022) · Frontiers in Psychiatry · doi:10.3389/fpsyt.2021.837149 · Cited by 67
“To this end, in this paper we propose an unsupervised distributed adaptation method to align data between different users, that is, Euclidean alignment (EA) (37).”
MethodFerrero, L., Quiles, V., Ortiz, M., Iáñez, E., Gil-Agudo, A. M., & Azorín, J. M. (2023) · iScience · doi:10.1016/j.isci.2023.106675 · Cited by 73
“However, the effectiveness of these approaches can vary across individuals.(36,37)”
BackgroundJin, W., Zhu, X., Qian, L., Wu, C., Yang, F., Zhan, D., Kang, Z., Luo, K., Meng, D., & Xu, G. (2024) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2024.1431815 · Cited by 75
“…research on decoding methods and signal processing algorithms, extracting useful information from EEG signals remains challenging due to the inherently low signal-to-noise ratio (SNR), particularly in the context of controlling multi-degree-of-freedom neuroprosthetic systems (He and Wu, 2020).”
BackgroundKo, W., Jeon, E., Jeong, S., Phyo, J., & Suk, H. (2021) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2021.643386 · Cited by 71
MI-DABAN: A dual-attention-based adversarial network for motor imagery classification
Li, H., Zhang, D., & Xie, J. (2022) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2022.106420 · Cited by 59
On statistic alignment for domain adaptation in structural health monitoring
Poole, J., Gardner, P. A., Dervilis, N., Bull, L. A., & Worden, K. (2022) · Structural Health Monitoring · doi:10.1177/14759217221110441 · Cited by 62
“proposed Euclidean alignment (EA), which aligns the means of the covariance matrices for 2D electroencephalogram features, and demonstrated that SA methods can be generalised to cases with multiple sources.(47) A related branch of DA concerns batch normalisation approaches, which align the means and standard deviation…”
Motor imagery EEG decoding using manifold embedded transfer learning
Cai, Y., She, Q., Ji, J., Ma, Y., Zhang, J., & Zhang, Y. (2022) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2022.109489 · Cited by 53
Chen, Y., Yang, R., Huang, M., Wang, Z., & Liu, X. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3191869 · Cited by 58
Bibri, S. E. (2023) · Smart Cities · doi:10.3390/smartcities6030065 · Cited by 69
Singh, A. S. G., Lal, S. P., & Guesgen, H. W. (2019) · Sensors · doi:10.3390/s19020379 · Cited by 56
“Then, Riemannian distance is used as pattern recognition metric for classification as it is invariant to any linear invertible transformation [28].”
BackgroundMethodGu, X., Cao, Z., Jolfaei, A., Xu, P., Wu, D., Jung, T.-P., & Lin, C.-T. (2021) · IEEE/ACM Transactions on Computational Biology and Bioinformatics · doi:10.1109/tcbb.2021.3052811 · Cited by 395
“lity of reducing the number of training templates for an SSVEP BCI. Inter- and intra-subject transfer learning is also applied to unsupervised conditions when no labelled data is available. He and Wu [67] presented a method to align EEG trails directly in the Euclidean space across different subjects to increase the s…”
MethodSelf-citationMDTL: A Novel and Model-Agnostic Transfer Learning Strategy for Cross-Subject Motor Imagery BCI
Li, A., Wang, Z., Zhao, X., Xu, T., Zhou, T., & Hu, H. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3259730 · Cited by 58
“He and Wu [21] proposed a Euclidean alignment (EA) method to align EEG data from different subjects into Euclidean space and then extract common Euclidean space and Riemannian space features.”
MethodSub-band target alignment common spatial pattern in brain-computer interface
Zhang, X., She, Q., Chen, Y., Kong, W., & Mei, C. (2021) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2021.106150 · Cited by 50
Small Sample Motor Imagery Classification Using Regularized Riemannian Features
Singh, A. S. G., Lal, S. P., & Guesgen, H. W. (2019) · IEEE Access · doi:10.1109/access.2019.2909058 · Cited by 50
“The Riemannian distance follows congruence invariance [21], meaning it remains unchanged under linear invertible transformation.”
BackgroundMulti-Source Transfer Learning for EEG Classification Based on Domain Adversarial Neural Network
Liu, D., Zhang, J., Wu, H., Liu, S., & Long, J. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3219418 · Cited by 53
“• EA-CSP-LDA: EA (Euclidean-space Alignment) [56] aligns the EEG trials from different subjects in the Euclidean space to make them more similar.”
BackgroundTransfer learning promotes acquisition of individual BCI skills
Kumar, S., Alawieh, H., Rácz, F. S., Fakhreddine, R., & Millán, J. D. R. (2024) · PNAS Nexus · doi:10.1093/pnasnexus/pgae076 · Cited by 61
“Adaptive recentering in common spatial pattern: Building upon previous research (58, 59), we have integrated the GR framework into the classical CSP method and evaluated it in a pseudo-online setting.”
MethodTowards Adaptive Classification using Riemannian Geometry approaches in Brain-Computer Interfaces
Kumar, S., Yger, F., & Lotte, F. (2019) · International Winter Conference on Brain-Computer Interface (BCI) · doi:10.1109/iww-bci.2019.8737349 · Cited by 64
Fu, R., Chen, Y.-F., Huang, Y., Chen, S., Duan, F., Li, J., Wu, J., Jiang, D., Gao, J., Gu, J. J., et al. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3174821 · Cited by 46
Automated EEG mega-analysis I: Spectral and amplitude characteristics across studies
Bigdely-Shamlo, N., Touryan, J., Ojeda, A., Kothe, C. A., Mullen, T. R., & Robbins, K. A. (2019) · NeuroImage · doi:10.1016/j.neuroimage.2019.116361 · Cited by 46
Deep Adversarial Domain Adaptation With Few-Shot Learning for Motor-Imagery Brain-Computer Interface
Phunruangsakao, C., Achanccaray, D., & Hayashibe, M. (2022) · IEEE Access · doi:10.1109/access.2022.3178100 · Cited by 47
Emotion Recognition via Multiscale Feature Fusion Network and Attention Mechanism
Jiang, Y., Xie, S., Xie, X., Cui, Y., & Tang, H. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3265688 · Cited by 45
“In order to obtain samples with similar distributions, all samples are calibrated using the Euclidean space data alignment [43] method before feeding into the network.”
MethodLiang, G., Cao, D., Wang, J., Zhang, Z., & Wu, Y. (2024) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2024.3382226 · Cited by 53
“We compared the transfer learning performance of three condition-based domain adaptation (DA) methods, including JDAO-Mix [42], DJDAN [43], and EA-CSP-LDA [44], and three fine-tuning-based methods, including DeepConvNet, MSFBCNN [45], and LSTM-MLP-T [46].”
MethodHuang, X., Xu, Y., Hua, J., Yi, W., Yin, H., Hu, R., & Wang, S. (2021) · Frontiers in Neuroscience · doi:10.3389/fnins.2021.733546 · Cited by 56
Roy, R. N., Hinss, M. F., Darmet, L., Ladouce, S., Jahanpour, E., Somon, B., Xu, X., Drougard, N., Dehais, F., & Lotte, F. (2022) · Frontiers in Neuroergonomics · doi:10.3389/fnrgo.2022.838342 · Cited by 63
Chen, P., Wang, H., Sun, X., Li, H., Grebogi, C., & Gao, Z. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3211881 · Cited by 43
“In the following experiments, the previous state-of-the-art algorithms are employed for comparison, including EA [18], transfer component analysis (TCA) [25], joint distribution adaptation (JDA) [26], BOTDA [7], DANN [28], CDAN [30] and DJDAN [22] methods.”
BackgroundMethodInfluentialGITGAN: Generative inter-subject transfer for EEG motor imagery analysis
Yin, K., Lim, E. Y., Lee, S., Shin, H.-B., & Lee, S. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.110015 · Cited by 44
Fang, W., Wu, D., Love, P. E. D., Ding, L., & Luo, H. (2022) · Advanced Engineering Informatics · doi:10.1016/j.aei.2022.101729 · Cited by 41
Self-citationLiu, W., Guo, C., & Gao, C. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.121612 · Cited by 55
Unsupervised multi-source domain adaptation via contrastive learning for EEG classification
Xu, C., Song, Y., Zheng, Q., Wang, Q., & Heng, P. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.125452 · Cited by 50
MI-DAGSC: A domain adaptation approach incorporating comprehensive information from MI-EEG signals
Zhang, D., Li, H., Xie, J., & Li, D. (2023) · Neural Networks · doi:10.1016/j.neunet.2023.08.008 · Cited by 37
Design a Novel BCI for Neurorehabilitation Using Concurrent LFP and EEG Features: A Case Study
Feng, Z., Sun, Y., Qian, L., Qi, Y., Wang, Y., Guan, C., & Sun, Y. (2021) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2021.3115799 · Cited by 39
Lightweight Source-Free Transfer for Privacy-Preserving Motor Imagery Classification
Zhang, W., & Wu, D. (2022) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2022.3193731 · Cited by 40
“1) Classical approaches for MI classification. a) CSP-LDA, which utilizes CSP (six filters) to extract log-variance features [56] and LDA clas-sifier [12] for classification. b) EA-CSP-LDA, which performs Euclidean alignment [16] on the EEG signals before CSP-LDA.”
BackgroundMethodSelf-citationWu, D., & Mendel, J. M. (2019) · IEEE Transactions on Fuzzy Systems · doi:10.1109/tfuzz.2019.2930022 · Cited by 39
Self-citationTransfer Learning Based on Optimal Transport for Motor Imagery Brain-Computer Interfaces
Peterson, V., Nieto, N., Wyser, D., Lambercy, O., Gassert, R., Milone, D. H., & Spies, R. D. (2021) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2021.3105912 · Cited by 43
Low dimensional subject representation based transfer learning in eeg decoding
Jeng, P.-Y., Wei, C.-S., Jung, T.-P., & Wang, L.-C. (2021) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2020.3025865 · Cited by 35
“First, the baseline-aligning-based transformation method aligns the features of data by removing subject-specific baselines so that the model can be trained from all the aligned data [24], [25], [28].”
MethodA Survey on Brain-Computer Interface-Inspired Communications: Opportunities and Challenges
Hu, H., Wang, Z., Zhao, X., Li, R., Li, A., Si, Y., Wang, J., Zhou, T., & Xu, T. (2024) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2024.3396847 · Cited by 41
A Spatiotemporal Neural Network Framework for EEG-Based Emotion Recognition in Depression Assessment
Novak, T. J., & Dvorak, P. M. (2025) · Journal of Medical Sciences and Interdisciplinary Research · doi:10.51847/a2pboyhjw1 · Cited by 34
Gao, Y., Liu, Y., She, Q., & Zhang, J. (2022) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2022.3218453 · Cited by 35
“He and Wu extended the RA to Euclidean alignment (EA) in Euclidean spaces [6].”
BackgroundImproving NeuCube spiking neural network for EEG-based pattern recognition using transfer learning
Wu, X., Feng, Y., Lou, S., Zheng, H. R., Hu, B., Hong, Z., & Tan, J. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.01.087 · Cited by 34
Multi-Source Decentralized Transfer for Privacy-Preserving BCIs
Zhang, W., Wang, Z., & Wu, D. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3207494 · Cited by 42
“• EA-CSP-LDA, which performs first Euclidean alignment [14] on the raw EEG data before CSP-LDA.”
BackgroundMethodSelf-citationWang, Y., Han, M., Peng, Y., Zhao, R., Fan, D., Meng, X., Xu, H. L., Niu, H., Cheng, J., & Liu, T. (2024) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2024.106046 · Cited by 37
Reducing Calibration Efforts in RSVP Tasks With Multi-Source Adversarial Domain Adaptation
Wei, W., Qiu, S., Ma, X., Li, D., Wang, B., & He, H. (2020) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2020.3023761 · Cited by 34
“He and Wu [30] promoted a data alignment method in Euclidean space.”
MethodSelf-citationDouble Stage Transfer Learning for Brain–Computer Interfaces
Gao, Y., Li, M., Peng, Y. F., Fang, F. Z., & Zhang, Y. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3241301 · Cited by 30
“He and Wu [17] proposed a new Euclidean alignment (EA) approach to align EEG trials directly in Euclidean space.”
BackgroundMethodInfluentialSelf-supervised contrastive learning for EEG-based cross-subject motor imagery recognition
Li, W., Li, H., Sun, X., Kang, H., An, S., Wang, G., & Gao, Z. (2024) · Journal of Neural Engineering · doi:10.1088/1741-2552/ad3986 · Cited by 43
“First, the Euclidean Alignment (EA) approach [50] is used in the PA stage to align the MI signals of source and target domains, which coarsely reduces the difference in edge distribution between domains.”
Liang, Z., Zheng, Z., Chen, W., Pei, Z., Wang, J., & Chen, J. (2024) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2024.106311 · Cited by 37
FB-CGANet: filter bank channel group attention network for multi-class motor imagery classification
Chen, J., Yi, W., Wang, D., Du, J., Fu, L., & Li, T. (2022) · Journal of Neural Engineering · doi:10.1088/1741-2552/ac4852 · Cited by 32
“[2] and EA + CSP [26] were used as the baseline methods.”
InfluentialWang, X., Yang, R., & Huang, M. (2022) · Sensors · doi:10.3390/s22062241 · Cited by 32
“Euclidean Space Data Alignment Based Preprocessing As a type of transfer learning method, EA can realize similar distribution between the source domain data and the target domain data after Euclidean space transformation [28].”
MethodT-TIME: Test-Time Information Maximization Ensemble for Plug-and-Play BCIs
Li, S., Wang, Z., Luo, H., Ding, L., & Wu, D. (2023) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2023.3303289 · Cited by 37
“For multi-channel EEG signals, Riemannian geometry-based approach [11] and Euclidean alignment (EA) [12] have demonstrated outstanding performance.”
BackgroundMethodInfluentialSelf-citationAdversarial attacks and defenses in physiological computing: a systematic review
Wu, D., Xu, J., Fang, W., Zhang, Y., Yang, L., Xu, X., Luo, H., & Yu, X. M. (2022) · National Science Open · doi:10.1360/nso/20220023 · Cited by 48
Self-citationXu, L., Xu, M., Ma, Z., Wang, K., Jung, T., & Ming, D. (2021) · Journal of Neural Engineering · doi:10.1088/1741-2552/ac1ed2 · Cited by 34
“In this strategy [33, 34], each trial was whitened by a reference matrix M ∈ R C × C for each subject.”
Xu, Y., Huang, X., & Lan, Q. (2021) · Frontiers in Neuroscience · doi:10.3389/fnins.2021.779231 · Cited by 36
“Since most classifiers are designed for the Euclidean space instead of the Riemannian manifold, Euclidean alignment-(EA-) based approaches extend RA-based approaches in the Euclidean space using the Euclidean mean as the reference matrix (He and Wu, 2020).”
BackgroundMethodPan, L., Wang, K., Xu, L., Sun, X., Yi, W., Xu, M., & Ming, D. (2023) · Journal of Neural Engineering · doi:10.1088/1741-2552/ad0a01 · Cited by 36
“In addition, we contrasted the traditional CSP algorithm without data pre-alignment and the CSP algorithm with EA data pre-alignment, marked as EA-CSP [40].”
Ahuja, C. K., & Sethia, D. (2024) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2024.1421922 · Cited by 41
“Similarly, ETL involves techniques like Non-Parametric Alignment (Gretton et al., 2012), or Adversarial Learning (He and Wu, 2019).”
MethodA Multi-Source Transfer Joint Matching Method for Inter-Subject Motor Imagery Decoding
Wei, F., Xu, X., Jia, T., Zhang, D., & Wu, X. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3243257 · Cited by 33
“EA is Euclidean alignment by task-state knowledge as a preprocessing step [38].”
BackgroundMethodCross-Channel Specific-Mutual Feature Transfer Learning for Motor Imagery EEG Signals Decoding
Li, D., Wang, J., Xu, J., Fang, X., & Ji, Y. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3269512 · Cited by 30
“Transfer learning [34], [35] is used to reduce the difference in data distribution between training and testing data.”
Supervised Contrastive Learning-Based Domain Generalization Network for Cross-Subject Motor Decoding
Zhi, H., Yu, T., Gu, Z., Lin, Z., Che, L., Li, Y., & Yu, Z. (2024) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2024.3432934 · Cited by 39
“In EEG-based BCI field, data alignment approaches such as Euclidean-space alignment (EA) [29] are relevant examples of DG.”
Wang, H., Chen, P., Zhang, M., Zhang, J., Sun, X., Li, M., Yang, X., & Gao, Z. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3243339 · Cited by 34
“Eight state-of-the-art methods were applied for performance comparison, including ShallowConvNet [40], EEGNet [29], EA-CSP [43], deep representation-based domain adaptation (DRDA) [44], MS-MDA [45], dynamic adversarial adaptation network (DAAN) [46], conditional domain adversarial network (CDAN)…”
MethodChannel reflection: Knowledge-driven data augmentation for EEG-based brain–computer interfaces
Wang, Z., Li, S., Luo, J., Liu, J., & Wu, D. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106351 · Cited by 35
Self-citationImproving Generalization of CNN-Based Motor-Imagery EEG Decoders via Dynamic Convolutions
Barmpas, K., Panagakis, Y., Bakas, S., Adamos, D. A., Laskaris, N., & Zafeiriou, S. P. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3265304 · Cited by 33
“…framework with two other commonly used transfer learning EEG techniques: 1) an adversarial approach, namely [40], that (similarly to our approach) does not use a calibration set and 2) Euclidean alignment [53] that projects data into a domain-invariant space but it uses all the trials of a subject.”
MethodToward Domain-Free Transformer for Generalized EEG Pre-Training
Kim, S.-J., Lee, D.-H., Kwak, H.-G., & Lee, S. (2024) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2024.3355434 · Cited by 29
“For example, He et al. [50] proposed the Euclidean-space alignment (EA) which could alleviate the issues from covariance shift between data.”
BackgroundA Strong and Simple Deep Learning Baseline for BCI Motor Imagery Decoding
Ouahidi, Y. E., Gripon, V., Pasdeloup, B., Bouallegue, G., Farrugia, N., & Lioi, G. (2024) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2024.3451010 · Cited by 45
“EA addresses this issue by projecting the data into a domain-invariant space based on the covariance matrix, ensuring consistency across subjects and sessions by making data more homogeneous and less scattered [43].”
BackgroundFederated Motor Imagery Classification for Privacy-Preserving Brain-Computer Interfaces
Jia, T., Meng, L., Li, S., Liu, J., & Wu, D. (2024) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2024.3457504 · Cited by 43
“We performed Euclidean alignment (EA) [41] for each subject.”
MethodSelf-citationArı, E., & Taçgın, E. (2024) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2024.106081 · Cited by 29
Wang, Y., Wang, J., Wang, W., Su, J., Bunterngchit, C., & Hou, Z. (2024) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2024.3474049 · Cited by 34
“Based on Riemannian alignment, He et al. [23] proposed Euclidean alignment (EA), which transformed the average covariance matrix of all subjects into a unit matrix, thus reducing the difference between different subjects to achieve better classification performance.”
Decoding Coordinated Directions of Bimanual Movements From EEG Signals
Zhang, M., Wu, J., Song, J., Fu, R., Ma, R., Jiang, Y., & Chen, Y.-F. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3220884 · Cited by 29
“Furthermore, future research also may consider the perspectives that participants can try MI tasks [48].”
BackgroundSmall Data Least-Squares Transformation (sd-LST) for Fast Calibration of SSVEP-Based BCIs
Bian, R., Wu, H., Liu, B., & Wu, D. (2022) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2022.3225878 · Cited by 25
“Transfer learning [19], [20], [21] is frequently used to reduce the calibration effort.”
BackgroundSelf-citationRecognition of EEG based on Improved Black Widow Algorithm optimized SVM
Huang, Q., Wang, C., Ye, Y., et al. (2022) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2022.104454 · Cited by 25
Metadata
- Authors
- He, H., & Wu, D.
- Affiliations
- Huazhong University of Science and Technology (CN)
- ORCID
- Venue
- IEEE Transactions on Biomedical Engineering
- Published
- 2020-02
- Volume
- 67
- Issue
- 2
- Pages
- 399-410
- ISSN
- 0018-9294 (ISSN-L)
- 1558-2531
- OpenAlex
- W2963727766
- Cited by
- 569
- References
- 52
- Citations by year
- 2026: 135 · 2025: 117 · 2024: 101 · 2023: 69 · 2022: 70 · 2021: 39 · 2020: 26
- FWCI
- 19.17
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Biomedical Engineering (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
- Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
- Physical Sciences › Computer Science › Human-Computer Interaction › Gaze Tracking and Assistive Technology
- Keywords
- brain-computer interface, EEG, transfer learning, data alignment, interindividual variability, motor imagery classification, unsupervised alignment, EEG preprocessing
- MeSH terms
- Machine Learning, Algorithms, Computer Simulation, Electroencephalography, Evoked Potentials, Humans, Imagination, Signal Processing, Computer-Assisted, Databases, Factual, Brain-Computer Interfaces
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.