BENDR: Using Transformers and a Contrastive Self-Supervised Learning Task to Learn From Massive Amounts of EEG Data

Deep neural networks (DNNs) used for brain–computer interface (BCI) classification are commonly expected to learn general features when trained across a variety of contexts, such that these features could be fine-tuned to specific contexts. While some success is found in such an approach, we suggest...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Frontiers in human neuroscience Ročník 15; s. 653659
Hlavní autori: Kostas, Demetres, Aroca-Ouellette, Stéphane, Rudzicz, Frank
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Frontiers Media S.A 23.06.2021
Predmet:
ISSN:1662-5161, 1662-5161
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Abstract Deep neural networks (DNNs) used for brain–computer interface (BCI) classification are commonly expected to learn general features when trained across a variety of contexts, such that these features could be fine-tuned to specific contexts. While some success is found in such an approach, we suggest that this interpretation is limited and an alternative would better leverage the newly (publicly) available massive electroencephalography (EEG) datasets. We consider how to adapt techniques and architectures used for language modeling (LM) that appear capable of ingesting awesome amounts of data toward the development of encephalography modeling with DNNs in the same vein. We specifically adapt an approach effectively used for automatic speech recognition, which similarly (to LMs) uses a self-supervised training objective to learn compressed representations of raw data signals. After adaptation to EEG, we find that a single pre-trained model is capable of modeling completely novel raw EEG sequences recorded with differing hardware, and different subjects performing different tasks. Furthermore, both the internal representations of this model and the entire architecture can be fine-tuned to a variety of downstream BCI and EEG classification tasks, outperforming prior work in more task-specific (sleep stage classification) self-supervision.
AbstractList Deep neural networks (DNNs) used for brain–computer interface (BCI) classification are commonly expected to learn general features when trained across a variety of contexts, such that these features could be fine-tuned to specific contexts. While some success is found in such an approach, we suggest that this interpretation is limited and an alternative would better leverage the newly (publicly) available massive electroencephalography (EEG) datasets. We consider how to adapt techniques and architectures used for language modeling (LM) that appear capable of ingesting awesome amounts of data toward the development of encephalography modeling with DNNs in the same vein. We specifically adapt an approach effectively used for automatic speech recognition, which similarly (to LMs) uses a self-supervised training objective to learn compressed representations of raw data signals. After adaptation to EEG, we find that a single pre-trained model is capable of modeling completely novel raw EEG sequences recorded with differing hardware, and different subjects performing different tasks. Furthermore, both the internal representations of this model and the entire architecture can be fine-tuned to a variety of downstream BCI and EEG classification tasks, outperforming prior work in more task-specific (sleep stage classification) self-supervision.
Deep neural networks (DNNs) used for brain–computer interface (BCI) classification are commonly expected to learn general features when trained across a variety of contexts, such that these features could be fine-tuned to specific contexts. While some success is found in such an approach, we suggest that this interpretation is limited and an alternative would better leverage the newly (publicly) available massive electroencephalography (EEG) datasets. We consider how to adapt techniques and architectures used for language modeling (LM) that appear capable of ingesting awesome amounts of data toward the development of encephalography modeling with DNNs in the same vein. We specifically adapt an approach effectively used for automatic speech recognition, which similarly (to LMs) uses a self-supervised training objective to learn compressed representations of raw data signals. After adaptation to EEG, we find that a single pre-trained model is capable of modeling completely novel raw EEG sequences recorded with differing hardware, and different subjects performing different tasks. Furthermore, both the internal representations of this model and the entire architecture can be fine-tuned to a variety of downstream BCI and EEG classification tasks, outperforming prior work in more task-specific (sleep stage classification) self-supervision.
Deep neural networks (DNNs) used for brain-computer interface (BCI) classification are commonly expected to learn general features when trained across a variety of contexts, such that these features could be fine-tuned to specific contexts. While some success is found in such an approach, we suggest that this interpretation is limited and an alternative would better leverage the newly (publicly) available massive electroencephalography (EEG) datasets. We consider how to adapt techniques and architectures used for language modeling (LM) that appear capable of ingesting awesome amounts of data toward the development of encephalography modeling with DNNs in the same vein. We specifically adapt an approach effectively used for automatic speech recognition, which similarly (to LMs) uses a self-supervised training objective to learn compressed representations of raw data signals. After adaptation to EEG, we find that a single pre-trained model is capable of modeling completely novel raw EEG sequences recorded with differing hardware, and different subjects performing different tasks. Furthermore, both the internal representations of this model and the entire architecture can be fine-tuned to a variety of downstream BCI and EEG classification tasks, outperforming prior work in more task-specific (sleep stage classification) self-supervision.Deep neural networks (DNNs) used for brain-computer interface (BCI) classification are commonly expected to learn general features when trained across a variety of contexts, such that these features could be fine-tuned to specific contexts. While some success is found in such an approach, we suggest that this interpretation is limited and an alternative would better leverage the newly (publicly) available massive electroencephalography (EEG) datasets. We consider how to adapt techniques and architectures used for language modeling (LM) that appear capable of ingesting awesome amounts of data toward the development of encephalography modeling with DNNs in the same vein. We specifically adapt an approach effectively used for automatic speech recognition, which similarly (to LMs) uses a self-supervised training objective to learn compressed representations of raw data signals. After adaptation to EEG, we find that a single pre-trained model is capable of modeling completely novel raw EEG sequences recorded with differing hardware, and different subjects performing different tasks. Furthermore, both the internal representations of this model and the entire architecture can be fine-tuned to a variety of downstream BCI and EEG classification tasks, outperforming prior work in more task-specific (sleep stage classification) self-supervision.
Author Kostas, Demetres
Aroca-Ouellette, Stéphane
Rudzicz, Frank
AuthorAffiliation 1 Department Computer Science, University of Toronto , Toronto, ON , Canada
2 Vector Institute for Artificial Intelligence , Toronto, ON , Canada
3 Li Ka Shing Knowledge Institute, St. Michael's Hospital , Toronto, ON , Canada
AuthorAffiliation_xml – name: 3 Li Ka Shing Knowledge Institute, St. Michael's Hospital , Toronto, ON , Canada
– name: 1 Department Computer Science, University of Toronto , Toronto, ON , Canada
– name: 2 Vector Institute for Artificial Intelligence , Toronto, ON , Canada
Author_xml – sequence: 1
  givenname: Demetres
  surname: Kostas
  fullname: Kostas, Demetres
– sequence: 2
  givenname: Stéphane
  surname: Aroca-Ouellette
  fullname: Aroca-Ouellette, Stéphane
– sequence: 3
  givenname: Frank
  surname: Rudzicz
  fullname: Rudzicz, Frank
BookMark eNp9kU1v1DAQhi1URD_gB3DzkUsWfyfLAalst6XSAhJbztYktrcuib21k5X49ySbClEOnGY0nvfx6H3P0UmIwSL0lpIF59XyvQv3Q7dghNGFklzJ5Qt0RpVihaSKnvzVn6LznB8IUUxJ-gqdcsFEJRk9Q_nT-uvV9w_4R_Zhh-8ShOxi6mzKGILBgFcx9Aly7w8Wb23riu2wt-ngszV4YyGFow7yT9zHeYCvU-zwF8h50lx2cQh9xtHh9foGX0EPr9FLB222b57qBdper-9Wn4vNt5vb1eWmaIRgfWFqQ6iroFECylIaKZlxjtQOQChKgJRLKsBIQcvGOEVUVVbE8sqOThjOL9DtTDURHvQ--Q7SLx3B6-Mgpp2G1Pumtbp2EgQ0pCwdF7Z2Fa_L0VUOspFLKtnI-jiz9kPdWdPYyZP2GfT5S_D3ehcPumLjpXI65t0TIMXHweZedz43tm0h2DhkzaQkirNKLMdVOq82KeacrPvzDSV6il0fY9dT7HqOfdSU_2ga30Pvj9n59j_K34PZtMw
CitedBy_id crossref_primary_10_1109_TCYB_2024_3410844
crossref_primary_10_1007_s11571_024_10162_5
crossref_primary_10_1016_j_bspc_2023_105488
crossref_primary_10_1088_1741_2552_adc5a3
crossref_primary_10_1109_TNSRE_2025_3595379
crossref_primary_10_1002_brx2_29
crossref_primary_10_1109_TNSRE_2024_3355434
crossref_primary_10_1016_j_dscb_2024_100121
crossref_primary_10_3390_molecules26072065
crossref_primary_10_1007_s10618_024_01043_w
crossref_primary_10_1038_s41551_025_01442_4
crossref_primary_10_3390_genes15111457
crossref_primary_10_1109_LGRS_2025_3558852
crossref_primary_10_3390_brainsci14070688
crossref_primary_10_1109_ACCESS_2023_3329678
crossref_primary_10_1016_j_artmed_2024_102900
crossref_primary_10_1016_j_asoc_2025_113563
crossref_primary_10_1088_1741_2552_ad8962
crossref_primary_10_1109_ACCESS_2023_3344531
crossref_primary_10_1109_TNSRE_2022_3194600
crossref_primary_10_3389_fnhum_2024_1426055
crossref_primary_10_3390_bioengineering11050418
crossref_primary_10_1088_1741_2552_ac857c
crossref_primary_10_1016_j_inffus_2024_102697
crossref_primary_10_1109_ACCESS_2025_3604528
crossref_primary_10_62762_CJIF_2024_876830
crossref_primary_10_1109_TAFFC_2023_3263907
crossref_primary_10_1109_JBHI_2022_3213171
crossref_primary_10_3389_fnhum_2024_1421922
crossref_primary_10_3389_fnhum_2022_898300
crossref_primary_10_3390_ijms26189178
crossref_primary_10_1007_s10489_022_04228_2
crossref_primary_10_1038_s41598_022_18502_3
crossref_primary_10_1007_s42979_022_01118_9
crossref_primary_10_1016_j_bspc_2025_108648
crossref_primary_10_1109_TNSRE_2022_3229330
crossref_primary_10_3389_fphys_2023_1196919
crossref_primary_10_1088_1741_2552_ad4743
crossref_primary_10_1109_ACCESS_2024_3509519
crossref_primary_10_1109_TIM_2025_3533618
crossref_primary_10_1109_TNSRE_2022_3192448
crossref_primary_10_1145_3597306
crossref_primary_10_1016_j_neucom_2024_127243
crossref_primary_10_1038_s41598_023_41653_w
crossref_primary_10_1016_j_knosys_2025_114199
crossref_primary_10_1016_j_bspc_2024_106131
crossref_primary_10_1111_coin_12659
crossref_primary_10_1109_TNSRE_2022_3230250
crossref_primary_10_1007_s00422_023_00967_8
crossref_primary_10_1109_TNSRE_2022_3199363
crossref_primary_10_1109_ACCESS_2024_3519297
crossref_primary_10_1109_ACCESS_2024_3404634
crossref_primary_10_1088_1741_2552_ad3eb5
crossref_primary_10_3390_brainsci15040382
crossref_primary_10_1038_s41551_022_00914_1
crossref_primary_10_1088_1741_2552_ac4f9a
crossref_primary_10_1109_ACCESS_2024_3459866
crossref_primary_10_3390_s25051293
crossref_primary_10_1109_TNSRE_2023_3275172
crossref_primary_10_1109_TAFFC_2022_3210441
crossref_primary_10_1109_ACCESS_2024_3394696
crossref_primary_10_1109_JSEN_2024_3468951
crossref_primary_10_3389_fninf_2023_1272791
crossref_primary_10_1111_nyas_15288
crossref_primary_10_1109_RBME_2023_3296938
crossref_primary_10_1038_s41598_023_35004_y
crossref_primary_10_1109_TIM_2023_3341114
crossref_primary_10_1016_j_neures_2024_06_003
crossref_primary_10_1109_JBHI_2024_3416897
crossref_primary_10_1016_j_neunet_2025_107816
crossref_primary_10_1109_ACCESS_2025_3580145
crossref_primary_10_1016_j_neunet_2024_106100
crossref_primary_10_1038_s44325_024_00027_5
crossref_primary_10_1109_ACCESS_2025_3545094
crossref_primary_10_1109_JBHI_2023_3304646
crossref_primary_10_1145_3736574
crossref_primary_10_1007_s11263_022_01713_6
crossref_primary_10_1002_hbm_26500
crossref_primary_10_1038_s41598_025_86294_3
crossref_primary_10_1109_ACCESS_2023_3294618
crossref_primary_10_1109_TNSRE_2025_3530110
crossref_primary_10_3390_s24030877
crossref_primary_10_1109_TAFFC_2024_3524418
crossref_primary_10_3389_frai_2022_807406
crossref_primary_10_1007_s10618_023_00948_2
crossref_primary_10_3389_frai_2025_1502504
crossref_primary_10_1016_j_eswa_2023_121734
crossref_primary_10_1016_j_cmpb_2021_106604
crossref_primary_10_1016_j_patcog_2024_110726
crossref_primary_10_1016_j_jobe_2023_108180
crossref_primary_10_1109_ACCESS_2023_3320561
crossref_primary_10_1016_j_knosys_2025_113345
crossref_primary_10_1109_ACCESS_2025_3605265
crossref_primary_10_1109_TNSRE_2023_3268751
crossref_primary_10_1016_j_heliyon_2024_e31485
crossref_primary_10_3389_fnins_2022_971829
crossref_primary_10_1145_3649448
Cites_doi 10.3390/s20072034
10.1007/s11263-019-01198-w
10.1109/CVPR.2009.5206848
10.1109/TBME.2017.2742541
10.3389/fnhum.2017.00334
10.1007/s10548-009-0121-6
10.3389/fnins.2012.00055
10.1016/j.neuroimage.2006.09.024
10.1073/pnas.1907373117
10.1109/TNNLS.2019.2900046
10.1161/01.cir.101.23.e215
10.1162/tacl_a_00300
10.1371/journal.pone.0216456
10.1109/MLSP.2019.8918693
10.1109/JBHI.2020.2967128
10.1016/j.jneumeth.2015.01.033
10.1016/j.neuroimage.2020.117021
10.1088/1741-2552/abc902
10.1038/nature14539
10.1088/1741-2552/aab2f2
10.1101/2020.12.17.423197
10.1038/s41598-019-38612-9
10.1109/ICASSP40776.2020.9054224
10.3389/fnins.2016.00196
10.1088/1741-2560/7/5/056006
10.1109/ACCESS.2019.2930958
10.1109/CVPR.2016.90
10.1088/1741-2552/ab260c
10.1109/TNSRE.2018.2813138
10.1155/2012/578295
10.1371/journal.pone.0207351
10.1016/j.eswa.2018.08.031
10.1007/978-3-642-24797-2
10.1088/1741-2552/abca18
10.1109/TBME.2004.827072
10.1109/ACCESS.2019.2919143
10.1088/1741-2552/aaf3f6
10.1002/hbm.23730
10.1109/10.867928
10.1038/s41591-018-0171-y
10.1088/1741-2552/aace8c
10.21437/Interspeech.2020-1228
10.1088/1741-2552/abb7a7
ContentType Journal Article
Copyright Copyright © 2021 Kostas, Aroca-Ouellette and Rudzicz.
Copyright © 2021 Kostas, Aroca-Ouellette and Rudzicz. 2021 Kostas, Aroca-Ouellette and Rudzicz
Copyright_xml – notice: Copyright © 2021 Kostas, Aroca-Ouellette and Rudzicz.
– notice: Copyright © 2021 Kostas, Aroca-Ouellette and Rudzicz. 2021 Kostas, Aroca-Ouellette and Rudzicz
DBID AAYXX
CITATION
7X8
5PM
DOA
DOI 10.3389/fnhum.2021.653659
DatabaseName CrossRef
MEDLINE - Academic
PubMed Central (Full Participant titles)
DOAJ Directory of Open Access Journals
DatabaseTitle CrossRef
MEDLINE - Academic
DatabaseTitleList CrossRef

MEDLINE - Academic

Database_xml – sequence: 1
  dbid: DOA
  name: DOAJ Directory of Open Access Journals
  url: https://www.doaj.org/
  sourceTypes: Open Website
– sequence: 2
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Anatomy & Physiology
EISSN 1662-5161
ExternalDocumentID oai_doaj_org_article_bf5a4ac077f34ebf83b72023a5c59152
PMC8261053
10_3389_fnhum_2021_653659
GrantInformation_xml – fundername: Natural Sciences and Engineering Research Council of Canada
  grantid: RGPIN 435874
– fundername: Electronics and Telecommunications Research Institute
  grantid: 20ZS1100, Core Technology Research for Self- Improving Integrated Artificial Intelligence System
GroupedDBID ---
29H
2WC
53G
5GY
5VS
88I
8FE
8FH
9T4
AAFWJ
AAYXX
ABIVO
ABUWG
ACGFO
ACGFS
ADBBV
ADRAZ
AEGXH
AENEX
AFFHD
AFKRA
AFPKN
AIAGR
ALMA_UNASSIGNED_HOLDINGS
AOIJS
AZQEC
BAWUL
BBNVY
BCNDV
BENPR
BHPHI
BPHCQ
CCPQU
CITATION
CS3
DIK
DU5
DWQXO
E3Z
EMOBN
F5P
GNUQQ
GROUPED_DOAJ
GX1
HCIFZ
HYE
KQ8
LK8
M2P
M48
M7P
M~E
O5R
O5S
OK1
OVT
PGMZT
PHGZM
PHGZT
PIMPY
PQGLB
PQQKQ
PROAC
RNS
RPM
TR2
7X8
PUEGO
5PM
ID FETCH-LOGICAL-c442t-dbd01f8ac64a775d552dff0bfaa4610a07914ad5417cdf6068780e38e365d33
IEDL.DBID DOA
ISICitedReferencesCount 133
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000670249100001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1662-5161
IngestDate Fri Oct 03 12:44:25 EDT 2025
Tue Nov 04 01:57:09 EST 2025
Thu Sep 04 17:15:44 EDT 2025
Sat Nov 29 05:51:59 EST 2025
Tue Nov 18 22:02:41 EST 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Language English
License This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c442t-dbd01f8ac64a775d552dff0bfaa4610a07914ad5417cdf6068780e38e365d33
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
This article was submitted to Brain-Computer Interfaces, a section of the journal Frontiers in Human Neuroscience
Edited by: Sung Chan Jun, Gwangju Institute of Science and Technology, South Korea
Reviewed by: Dalin Zhang, Aalborg University, Denmark; Tomasz Maciej Rutkowski, RIKEN Center for Advanced Intelligence Project (AIP), Japan
OpenAccessLink https://doaj.org/article/bf5a4ac077f34ebf83b72023a5c59152
PMID 34248521
PQID 2550632849
PQPubID 23479
ParticipantIDs doaj_primary_oai_doaj_org_article_bf5a4ac077f34ebf83b72023a5c59152
pubmedcentral_primary_oai_pubmedcentral_nih_gov_8261053
proquest_miscellaneous_2550632849
crossref_primary_10_3389_fnhum_2021_653659
crossref_citationtrail_10_3389_fnhum_2021_653659
PublicationCentury 2000
PublicationDate 2021-06-23
PublicationDateYYYYMMDD 2021-06-23
PublicationDate_xml – month: 06
  year: 2021
  text: 2021-06-23
  day: 23
PublicationDecade 2020
PublicationTitle Frontiers in human neuroscience
PublicationYear 2021
Publisher Frontiers Media S.A
Publisher_xml – name: Frontiers Media S.A
References Roy (B60) 2019; 16
B23
B69
B27
Baevski (B4) 2020
Huh (B34) 2016
Zhang (B76); 18
Arora (B3) 2019
He (B28) 2019
Fahimi (B22) 2019; 16
Citi (B15) 2010; 7
Lin (B48) 2017; 11
Cimtay (B14) 2020; 20
Kemp (B38) 2018
Chen (B11) 2016
Chambon (B10) 2018; 26
B30
Tang (B66) 2020
Chen (B12) 2020
van den Oord (B68) 2018
Kemp (B39) 2000; 47
Mohamed (B51) 2019
Raghu (B57) 2019
Sejnowski (B65) 2020; 117
Wu (B71) 2020; 128
Jurcak (B37) 2007; 34
Lotte (B49) 2018; 15
Schirrmeister (B63) 2017; 38
Yosinski (B73) 2015
Dose (B20) 2018; 114
B2
Jiang (B35) 2020
B5
Ravanelli (B58) 2018
Vidaurre (B70) 2010; 23
B8
B9
Ahn (B1) 2015; 243
Rivest (B59) 2020; 31
Sannelli (B61) 2019; 14
Kornblith (B41) 2019
Banville (B6) 2019
Deng (B17) 2009
Gemein (B24) 2020; 220
Kostas (B44); 17
Margaux (B50) 2012; 2012
Kostas (B43)
Tangermann (B67) 2012; 6
Zhang (B75); 24
Devlin (B18) 2019
Ditthapron (B19) 2019; 7
He (B29) 2016
LeCun (B47) 2015; 521
Zanini (B74) 2018; 65
Huang (B32) 2017
Goldberger (B25) 2000; 101
Krizhevsky (B45) 2012
Ngiam (B54) 2018
Mousavi (B52) 2019; 14
Huang (B33) 2020
B53
Kostas (B42) 2019; 9
Kingma (B40) 2015
Raffel (B56) 2020
Lawhern (B46) 2018; 15
Citi (B16) 2014
Schalk (B62) 2004; 51
Obeid (B55) 2016; 10
Schwemmer (B64) 2018; 24
Dosovitskiy (B21) 2020
Xu (B72) 2019; 7
Chung (B13) 2020
Hendrycks (B31) 2016
Banville (B7) 2020; 18
Graves (B26) 2012
Joshi (B36) 2020; 8
References_xml – volume: 20
  start-page: 1
  year: 2020
  ident: B14
  article-title: Investigating the use of pretrained convolutional neural network on cross-subject and cross-dataset eeg emotion recognition
  publication-title: Sensors
  doi: 10.3390/s20072034
– start-page: 4171
  volume-title: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019
  year: 2019
  ident: B18
  article-title: BERT: pre-training of deep bidirectional transformers for language understanding,
– start-page: 1
  volume-title: arXiv
  year: 2020
  ident: B12
  article-title: Big self-supervised models are strong semi-supervised learners
– volume: 128
  start-page: 742
  year: 2020
  ident: B71
  article-title: Group normalization
  publication-title: Int. J. Comput. Vis.
  doi: 10.1007/s11263-019-01198-w
– volume-title: CVPR09
  year: 2009
  ident: B17
  article-title: ImageNet: a large-scale hierarchical image database,
  doi: 10.1109/CVPR.2009.5206848
– volume: 65
  start-page: 1107
  year: 2018
  ident: B74
  article-title: Transfer learning: a riemannian geometry framework with applications to brain-computer interfaces
  publication-title: IEEE Trans. Biomed. Eng.
  doi: 10.1109/TBME.2017.2742541
– volume: 11
  start-page: 334
  year: 2017
  ident: B48
  article-title: Improving EEG-based emotion classification using conditional transfer learning
  publication-title: Front. Hum. Neurosci.
  doi: 10.3389/fnhum.2017.00334
– volume: 23
  start-page: 194
  year: 2010
  ident: B70
  article-title: Towards a cure for BCI illiteracy
  publication-title: Brain Topography
  doi: 10.1007/s10548-009-0121-6
– volume-title: arXiv
  year: 2016
  ident: B11
  article-title: ABC-CNN: an attention based convolutional neural network for visual question answering
– volume: 6
  start-page: 55
  year: 2012
  ident: B67
  article-title: Review of the BCI competition IV
  publication-title: Front. Neurosci.
  doi: 10.3389/fnins.2012.00055
– start-page: 9868
  volume-title: Proceedings of Machine Learning and Systems 2020
  year: 2020
  ident: B33
  article-title: Improving transformer optimization through better initialization,
– year: 2014
  ident: B16
  article-title: Erp-based brain-computer interface recordings
– volume: 34
  start-page: 1600
  year: 2007
  ident: B37
  article-title: 10/20, 10/10, and 10/5 systems revisited: their validity as relative head-surface-based positioning systems
  publication-title: Neuroimage
  doi: 10.1016/j.neuroimage.2006.09.024
– ident: B2
– volume: 117
  start-page: 30033
  year: 2020
  ident: B65
  article-title: The unreasonable effectiveness of deep learning in artificial intelligence
  publication-title: Proc. Natl. Acad. Sci. U.S.A.
  doi: 10.1073/pnas.1907373117
– volume: 31
  start-page: 174
  year: 2020
  ident: B59
  article-title: A new timing error cost function for binary time series prediction
  publication-title: IEEE Trans. Neural Netw. Learn. Syst.
  doi: 10.1109/TNNLS.2019.2900046
– ident: B69
– volume: 101
  start-page: E215
  year: 2000
  ident: B25
  article-title: PhysioBank, physioToolkit, and physioNet: components of a new research resource for complex physiologic signals
  publication-title: Circulation
  doi: 10.1161/01.cir.101.23.e215
– volume: 8
  start-page: 64
  year: 2020
  ident: B36
  article-title: SpanBERT: improving pre-training by representing and predicting spans
  publication-title: Trans. Assoc. Comput. Linguist.
  doi: 10.1162/tacl_a_00300
– start-page: 1
  year: 2020
  ident: B66
  article-title: Long-tailed classification by keeping the good and removing the bad momentum causal effect
  publication-title: NeurIPS
– volume: 14
  start-page: e0216456
  year: 2019
  ident: B52
  article-title: SleepEEGNet: automated sleep stage scoring with sequence to sequence deep learning approach
  publication-title: PLoS ONE
  doi: 10.1371/journal.pone.0216456
– volume-title: arXiv arXiv:2010.11929
  year: 2020
  ident: B21
  article-title: An image is worth 16x16 words: transformers for image recognition at scale
– start-page: 1
  volume-title: 2019 IEEE 29th International Workshop on Machine Learning for Signal Processing (MLSP)
  year: 2019
  ident: B6
  article-title: Self-supervised representation learning from electroencephalography signals,
  doi: 10.1109/MLSP.2019.8918693
– volume: 24
  start-page: 2570
  ident: B75
  article-title: Motor imagery classification via temporal attention cues of graph embedded EEG signals
  publication-title: IEEE J. Biomed. Health Informat.
  doi: 10.1109/JBHI.2020.2967128
– volume: 243
  start-page: 103
  year: 2015
  ident: B1
  article-title: Performance variation in motor imagery brain-computer interface: a brief review
  publication-title: J. Neurosci. Methods
  doi: 10.1016/j.jneumeth.2015.01.033
– volume: 220
  start-page: 17021
  year: 2020
  ident: B24
  article-title: Machine-learning-based diagnostics of EEG pathology
  publication-title: Neuroimage
  doi: 10.1016/j.neuroimage.2020.117021
– ident: B30
– volume: 18
  start-page: 031002
  ident: B76
  article-title: A survey on deep learning-based non-invasive brain signals: recent advances and new frontiers
  publication-title: J. Neural Eng
  doi: 10.1088/1741-2552/abc902
– volume: 521
  start-page: 436
  year: 2015
  ident: B47
  article-title: Deep learning
  publication-title: Nature
  doi: 10.1038/nature14539
– volume: 15
  start-page: 031005
  year: 2018
  ident: B49
  article-title: A review of classification algorithms for EEG-based brain-computer interfaces: a 10 year update
  publication-title: J. Neural Eng.
  doi: 10.1088/1741-2552/aab2f2
– ident: B9
– volume-title: bioRxiv
  ident: B43
  article-title: Dn3: an open-source python library for large-scale raw neurophysiology data assimilation for more flexible and standardized deep learning
  doi: 10.1101/2020.12.17.423197
– volume: 9
  start-page: 1609
  year: 2019
  ident: B42
  article-title: Machine learning for MEG during speech tasks
  publication-title: Sci. Rep.
  doi: 10.1038/s41598-019-38612-9
– start-page: 7694
  volume-title: ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
  year: 2020
  ident: B4
  article-title: Effectiveness of self-supervised pre-training for ASR,
  doi: 10.1109/ICASSP40776.2020.9054224
– volume: 10
  start-page: 196
  year: 2016
  ident: B55
  article-title: The temple university hospital EEG data corpus
  publication-title: Front. Neurosci.
  doi: 10.3389/fnins.2016.00196
– volume: 7
  start-page: 056006
  year: 2010
  ident: B15
  article-title: Documenting, modelling and exploiting P300 amplitude changes due to variable target delays in Donchins speller
  publication-title: J. Neural Eng.
  doi: 10.1088/1741-2560/7/5/056006
– start-page: 9904
  year: 2019
  ident: B3
  article-title: A theoretical analysis of contrastive unsupervised representation learning,
  publication-title: 36th International Conference on Machine Learning, ICML 2019
– volume: 7
  start-page: 112767
  year: 2019
  ident: B72
  article-title: A deep transfer convolutional neural network framework for EEG signal classification
  publication-title: IEEE Access
  doi: 10.1109/ACCESS.2019.2930958
– start-page: 770
  volume-title: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2016
  year: 2016
  ident: B29
  article-title: Deep residual learning for image recognition,
  doi: 10.1109/CVPR.2016.90
– year: 2019
  ident: B57
  article-title: Transfusion: Understanding transfer learning for medical imaging
  publication-title: arXiv
– volume: 16
  start-page: 051001
  year: 2019
  ident: B60
  article-title: Deep learning-based electroencephalography analysis: a systematic review
  publication-title: J. Neural Eng.
  doi: 10.1088/1741-2552/ab260c
– volume: 26
  start-page: 758
  year: 2018
  ident: B10
  article-title: A deep learning architecture for temporal sleep stage classification using multivariate and multimodal time series
  publication-title: IEEE Trans. Neural Syst. Rehabil. Eng.
  doi: 10.1109/TNSRE.2018.2813138
– ident: B27
– ident: B5
– start-page: 1097
  volume-title: Proceedings of the 25th International Conference on Neural Information Processing Systems
  year: 2012
  ident: B45
  article-title: ImageNet classification with deep convolutional neural Networks,
– volume: 2012
  start-page: 1
  year: 2012
  ident: B50
  article-title: Objective and subjective evaluation of online error correction during P300-Based spelling
  publication-title: Adv. Hum. Comput. Interact.
  doi: 10.1155/2012/578295
– volume-title: arXiv arXiv:1606.08415.
  year: 2016
  ident: B31
  article-title: Bridging nonlinearities and stochastic regularizers with Gaussian error linear units
– year: 2018
  ident: B38
  article-title: The sleep-edf database [expanded]
– volume: 14
  start-page: e0207351
  year: 2019
  ident: B61
  article-title: A large scale screening study with a SMR-based BCI: categorization of BCI users and differences in their SMR activity
  publication-title: PLoS ONE
  doi: 10.1371/journal.pone.0207351
– ident: B23
– start-page: 2656
  volume-title: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
  year: 2019
  ident: B41
  article-title: Do better imagenet models transfer better?
– year: 2018
  ident: B54
  article-title: Domain adaptive transfer learning with specialist models
  publication-title: arXiv
– volume: 114
  start-page: 532
  year: 2018
  ident: B20
  article-title: An end-to-end deep learning approach to MI-EEG signal classification for BCIs
  publication-title: Exp. Syst. Appl.
  doi: 10.1016/j.eswa.2018.08.031
– year: 2018
  ident: B58
  article-title: Interpretable convolutional filters with sincNet
  publication-title: Arxiv
– volume-title: Supervised Sequence Labelling with Recurrent Neural Networks
  year: 2012
  ident: B26
  doi: 10.1007/978-3-642-24797-2
– volume: 18
  start-page: 046020
  year: 2020
  ident: B7
  article-title: Uncovering the structure of clinical EEG signals with self-supervised learning
  publication-title: J. Neural Eng.
  doi: 10.1088/1741-2552/abca18
– volume-title: arXiv
  year: 2019
  ident: B51
  article-title: Transformers with convolutional context for ASR
– year: 2015
  ident: B73
  article-title: Understanding neural networks through deep visualization
  publication-title: arXiv arXiv:1506.06579.
– volume-title: arXiv arXiv:2005.09862.
  year: 2020
  ident: B35
  article-title: A further study of unsupervised pre-training for transformer based speech recognition
– volume: 51
  start-page: 1034
  year: 2004
  ident: B62
  article-title: BCI2000 : a general-purpose brain-computer interface (BCI) system
  publication-title: IEEE Trans. Biomed. Eng.
  doi: 10.1109/TBME.2004.827072
– start-page: 1
  volume-title: 3rd International Conference on Learning Representations, ICLR 2015-Conference Track Proceedings
  year: 2015
  ident: B40
  article-title: Adam: a method for stochastic optimization,
– ident: B8
– start-page: 1
  volume-title: CoRR
  year: 2016
  ident: B34
  article-title: What makes imageNet good for transfer learning?
– year: 2018
  ident: B68
  article-title: Representation learning with contrastive predictive coding
  publication-title: arXiv arXiv:1807.03748.
– volume: 7
  start-page: 68415
  year: 2019
  ident: B19
  article-title: Universal joint feature extraction for P300 EEG classification using multi-task autoencoder
  publication-title: IEEE Access
  doi: 10.1109/ACCESS.2019.2919143
– start-page: 2261
  volume-title: 2017 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017
  year: 2017
  ident: B32
  article-title: Densely connected convolutional networks,
– volume: 16
  start-page: 026007
  year: 2019
  ident: B22
  article-title: Inter-subject transfer learning with an end-to-end deep convolutional neural network for EEG-based BCI
  publication-title: J. Neural Eng.
  doi: 10.1088/1741-2552/aaf3f6
– volume: 38
  start-page: 5391
  year: 2017
  ident: B63
  article-title: Deep learning with convolutional neural networks for EEG decoding and visualization
  publication-title: Hum. Brain Mapp.
  doi: 10.1002/hbm.23730
– start-page: 4917
  volume-title: Proceedings of the IEEE International Conference on Computer Vision
  year: 2019
  ident: B28
  article-title: Rethinking imageNet pre-training,
– start-page: 1
  year: 2020
  ident: B56
  article-title: Exploring the limits of transfer learning with a unified text-to-text transformer
  publication-title: J. Mach. Learn. Res.
– volume: 47
  start-page: 1185
  year: 2000
  ident: B39
  article-title: Analysis of a sleep-dependent neuronal feedback loop: the slow-wave microcontinuity of the EEG
  publication-title: IEEE Trans. Biomed. Eng.
  doi: 10.1109/10.867928
– volume: 24
  start-page: 1669
  year: 2018
  ident: B64
  article-title: Meeting brain-computer interface user performance expectations using a deep neural network decoding framework
  publication-title: Nat. Med.
  doi: 10.1038/s41591-018-0171-y
– volume: 15
  start-page: aace8c
  year: 2018
  ident: B46
  article-title: EEGNet: a compact convolutional neural network for EEG-based brain-computer interfaces
  publication-title: J. Neural Eng.
  doi: 10.1088/1741-2552/aace8c
– start-page: 3760
  volume-title: Interspeech 2020
  year: 2020
  ident: B13
  article-title: Vector-quantized autoregressive predictive coding,
  doi: 10.21437/Interspeech.2020-1228
– volume: 17
  start-page: 56008
  ident: B44
  article-title: Thinker invariance: enabling deep neural networks for BCI across more people
  publication-title: J. Neural Eng.
  doi: 10.1088/1741-2552/abb7a7
– ident: B53
SSID ssj0062651
Score 2.6681752
Snippet Deep neural networks (DNNs) used for brain–computer interface (BCI) classification are commonly expected to learn general features when trained across a...
Deep neural networks (DNNs) used for brain-computer interface (BCI) classification are commonly expected to learn general features when trained across a...
SourceID doaj
pubmedcentral
proquest
crossref
SourceType Open Website
Open Access Repository
Aggregation Database
Enrichment Source
Index Database
StartPage 653659
SubjectTerms brain computer interface
contrastive learning
convolutional neural network
deep learning - artificial neural network
Human Neuroscience
semi-supervised learning
transformers
Title BENDR: Using Transformers and a Contrastive Self-Supervised Learning Task to Learn From Massive Amounts of EEG Data
URI https://www.proquest.com/docview/2550632849
https://pubmed.ncbi.nlm.nih.gov/PMC8261053
https://doaj.org/article/bf5a4ac077f34ebf83b72023a5c59152
Volume 15
WOSCitedRecordID wos000670249100001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVAON
  databaseName: DOAJ Directory of Open Access Journals
  customDbUrl:
  eissn: 1662-5161
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0062651
  issn: 1662-5161
  databaseCode: DOA
  dateStart: 20080101
  isFulltext: true
  titleUrlDefault: https://www.doaj.org/
  providerName: Directory of Open Access Journals
– providerCode: PRVHPJ
  databaseName: ROAD: Directory of Open Access Scholarly Resources
  customDbUrl:
  eissn: 1662-5161
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0062651
  issn: 1662-5161
  databaseCode: M~E
  dateStart: 20080101
  isFulltext: true
  titleUrlDefault: https://road.issn.org
  providerName: ISSN International Centre
– providerCode: PRVPQU
  databaseName: Biological Science Database
  customDbUrl:
  eissn: 1662-5161
  dateEnd: 20211231
  omitProxy: false
  ssIdentifier: ssj0062651
  issn: 1662-5161
  databaseCode: M7P
  dateStart: 20080328
  isFulltext: true
  titleUrlDefault: http://search.proquest.com/biologicalscijournals
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: ProQuest Central
  customDbUrl:
  eissn: 1662-5161
  dateEnd: 20211231
  omitProxy: false
  ssIdentifier: ssj0062651
  issn: 1662-5161
  databaseCode: BENPR
  dateStart: 20080328
  isFulltext: true
  titleUrlDefault: https://www.proquest.com/central
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: Publicly Available Content Database
  customDbUrl:
  eissn: 1662-5161
  dateEnd: 20211231
  omitProxy: false
  ssIdentifier: ssj0062651
  issn: 1662-5161
  databaseCode: PIMPY
  dateStart: 20080328
  isFulltext: true
  titleUrlDefault: http://search.proquest.com/publiccontent
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: Science Database
  customDbUrl:
  eissn: 1662-5161
  dateEnd: 20211231
  omitProxy: false
  ssIdentifier: ssj0062651
  issn: 1662-5161
  databaseCode: M2P
  dateStart: 20080328
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/sciencejournals
  providerName: ProQuest
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV3Nb9MwFLdgcOCCBgNRGJWREAeksCS2Y4dbx1JAolW0TqicIscf2rQ1mZoUiQt_O-8l6dRe4MLFB3_Ijt_L8-_FL79HyNso8S4sUx4YZkzAvXNBWnITOG5VCvg29V0ymO_f5Hyulss030n1hTFhPT1wv3EnpReaaxNK6Rl3pVeslJjyWwsjUjh80PoC6tk6U70NBpQuov4OE1yw9MRXlxv87TyOPiSCJUhMunMKdWT9ewhzPz5y58CZHpLHA1Kkk36FT8g9Vz0lR5MKvOTVL_qOdrGb3UfxI9KcZvOz84-0CwCgF1swCtCO6spSTZGEaq0btG104W58sNjcopVonKUDxSqM0801beu-gk7X9YrOAFvjmMkKc0o0tPY0yz7TM93qZ2QxzS4-fQmGdAqB4TxuA1vaMPJKm4RrKYUVIrbeh6XXGknXdSjTiGsreCSN9eDYKKlCx5SDDbOMPScHVV25F4QyLjzAPukSwbm3cSmUEyqSUWRTBgBkRMLt5hZmYBrHhBc3BXgcKI-ik0eB8ih6eYzI-7shtz3Nxt86n6LE7joiQ3ZXAXpTDHpT_EtvRuTNVt4FvFF4TaIrV2-aApwswG1wbMNEck8R9mbcb6muLjtubvDWALGyl_9jia_II3xqDEyL2TE5aNcb95o8ND_bq2Y9JvflUo3JA1Cw_HzcqT-UszjHUnbl7wza86-z_McfJ6MOzg
linkProvider Directory of Open Access Journals
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=BENDR%3A+Using+Transformers+and+a+Contrastive+Self-Supervised+Learning+Task+to+Learn+From+Massive+Amounts+of+EEG+Data&rft.jtitle=Frontiers+in+human+neuroscience&rft.au=Kostas%2C+Demetres&rft.au=Aroca-Ouellette%2C+St%C3%A9phane&rft.au=Rudzicz%2C+Frank&rft.date=2021-06-23&rft.pub=Frontiers+Media+S.A&rft.eissn=1662-5161&rft.volume=15&rft_id=info:doi/10.3389%2Ffnhum.2021.653659&rft_id=info%3Apmid%2F34248521&rft.externalDocID=PMC8261053
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1662-5161&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1662-5161&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1662-5161&client=summon