CapMatch: Semi-Supervised Contrastive Transformer Capsule With Feature-Based Knowledge Distillation for Human Activity Recognition

This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervise...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	IEEE transaction on neural networks and learning systems Ročník 36; číslo 2; s. 2690 - 2704
Hlavní autoři:	Xiao, Zhiwen, Tong, Huagang, Qu, Rong, Xing, Huanlai, Luo, Shouxi, Zhu, Zonghai, Song, Fuhong, Feng, Li
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	United States IEEE 01.02.2025
Témata:	Algorithms Capsule network (CapNet) Classification algorithms contrastive learning (CL) Data mining Feature extraction Human Activities - classification Human activity recognition human activity recognition (HAR) Humans knowledge distillation (KD) Neural Networks, Computer Pattern Recognition, Automated - methods Semantics semi-supervised learning (SSL) similarity learning Supervised Machine Learning Transformers Unsupervised learning Unsupervised Machine Learning Wearable Electronic Devices wearable sensors
ISSN:	2162-237X, 2162-2388, 2162-2388
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Abstract	This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervised learning and unsupervised learning to extract rich representations from input data. In unsupervised learning, CapMatch leverages the pseudolabeling, contrastive learning (CL), and feature-based KD techniques to construct similarity learning on lower and higher level semantic information extracted from two augmentation versions of the data, "weak" and "timecut," to recognize the relationships among the obtained features of classes in the unlabeled data. CapMatch combines the outputs of the weak- and timecut-augmented models to form pseudolabeling and thus CL. Meanwhile, CapMatch uses the feature-based KD to transfer knowledge from the intermediate layers of the weak-augmented model to those of the timecut-augmented model. To effectively capture both local and global patterns of HAR data, we design a capsule transformer network consisting of four capsule-based transformer blocks and one routing layer. Experimental results show that compared with a number of state-of-the-art semi-supervised and supervised algorithms, the proposed CapMatch achieves decent performance on three commonly used HAR datasets, namely, HAPT, WISDM, and UCI_HAR. With only 10% of data labeled, CapMatch achieves <inline-formula> <tex-math notation="LaTeX">F_{1} </tex-math></inline-formula> values of higher than 85.00% on these datasets, outperforming 14 semi-supervised algorithms. When the proportion of labeled data reaches 30%, CapMatch obtains <inline-formula> <tex-math notation="LaTeX">F_{1} </tex-math></inline-formula> values of no lower than 88.00% on the datasets above, which is better than several classical supervised algorithms, e.g., decision tree and <inline-formula> <tex-math notation="LaTeX">k </tex-math></inline-formula>-nearest neighbor (KNN).
AbstractList	This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervised learning and unsupervised learning to extract rich representations from input data. In unsupervised learning, CapMatch leverages the pseudolabeling, contrastive learning (CL), and feature-based KD techniques to construct similarity learning on lower and higher level semantic information extracted from two augmentation versions of the data, "weak" and "timecut," to recognize the relationships among the obtained features of classes in the unlabeled data. CapMatch combines the outputs of the weak- and timecut-augmented models to form pseudolabeling and thus CL. Meanwhile, CapMatch uses the feature-based KD to transfer knowledge from the intermediate layers of the weak-augmented model to those of the timecut-augmented model. To effectively capture both local and global patterns of HAR data, we design a capsule transformer network consisting of four capsule-based transformer blocks and one routing layer. Experimental results show that compared with a number of state-of-the-art semi-supervised and supervised algorithms, the proposed CapMatch achieves decent performance on three commonly used HAR datasets, namely, HAPT, WISDM, and UCI_HAR. With only 10% of data labeled, CapMatch achieves <inline-formula> <tex-math notation="LaTeX">F_{1} </tex-math></inline-formula> values of higher than 85.00% on these datasets, outperforming 14 semi-supervised algorithms. When the proportion of labeled data reaches 30%, CapMatch obtains <inline-formula> <tex-math notation="LaTeX">F_{1} </tex-math></inline-formula> values of no lower than 88.00% on the datasets above, which is better than several classical supervised algorithms, e.g., decision tree and <inline-formula> <tex-math notation="LaTeX">k </tex-math></inline-formula>-nearest neighbor (KNN). This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervised learning and unsupervised learning to extract rich representations from input data. In unsupervised learning, CapMatch leverages the pseudolabeling, contrastive learning (CL), and feature-based KD techniques to construct similarity learning on lower and higher level semantic information extracted from two augmentation versions of the data, "weak" and "timecut," to recognize the relationships among the obtained features of classes in the unlabeled data. CapMatch combines the outputs of the weak- and timecut-augmented models to form pseudolabeling and thus CL. Meanwhile, CapMatch uses the feature-based KD to transfer knowledge from the intermediate layers of the weak-augmented model to those of the timecut-augmented model. To effectively capture both local and global patterns of HAR data, we design a capsule transformer network consisting of four capsule-based transformer blocks and one routing layer. Experimental results show that compared with a number of state-of-the-art semi-supervised and supervised algorithms, the proposed CapMatch achieves decent performance on three commonly used HAR datasets, namely, HAPT, WISDM, and UCI_HAR. With only 10% of data labeled, CapMatch achieves values of higher than 85.00% on these datasets, outperforming 14 semi-supervised algorithms. When the proportion of labeled data reaches 30%, CapMatch obtains values of no lower than 88.00% on the datasets above, which is better than several classical supervised algorithms, e.g., decision tree and -nearest neighbor (KNN). This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervised learning and unsupervised learning to extract rich representations from input data. In unsupervised learning, CapMatch leverages the pseudolabeling, contrastive learning (CL), and feature-based KD techniques to construct similarity learning on lower and higher level semantic information extracted from two augmentation versions of the data, "weak" and "timecut," to recognize the relationships among the obtained features of classes in the unlabeled data. CapMatch combines the outputs of the weak- and timecut-augmented models to form pseudolabeling and thus CL. Meanwhile, CapMatch uses the feature-based KD to transfer knowledge from the intermediate layers of the weak-augmented model to those of the timecut-augmented model. To effectively capture both local and global patterns of HAR data, we design a capsule transformer network consisting of four capsule-based transformer blocks and one routing layer. Experimental results show that compared with a number of state-of-the-art semi-supervised and supervised algorithms, the proposed CapMatch achieves decent performance on three commonly used HAR datasets, namely, HAPT, WISDM, and UCI_HAR. With only 10% of data labeled, CapMatch achieves values of higher than 85.00% on these datasets, outperforming 14 semi-supervised algorithms. When the proportion of labeled data reaches 30%, CapMatch obtains values of no lower than 88.00% on the datasets above, which is better than several classical supervised algorithms, e.g., decision tree and -nearest neighbor (KNN).This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervised learning and unsupervised learning to extract rich representations from input data. In unsupervised learning, CapMatch leverages the pseudolabeling, contrastive learning (CL), and feature-based KD techniques to construct similarity learning on lower and higher level semantic information extracted from two augmentation versions of the data, "weak" and "timecut," to recognize the relationships among the obtained features of classes in the unlabeled data. CapMatch combines the outputs of the weak- and timecut-augmented models to form pseudolabeling and thus CL. Meanwhile, CapMatch uses the feature-based KD to transfer knowledge from the intermediate layers of the weak-augmented model to those of the timecut-augmented model. To effectively capture both local and global patterns of HAR data, we design a capsule transformer network consisting of four capsule-based transformer blocks and one routing layer. Experimental results show that compared with a number of state-of-the-art semi-supervised and supervised algorithms, the proposed CapMatch achieves decent performance on three commonly used HAR datasets, namely, HAPT, WISDM, and UCI_HAR. With only 10% of data labeled, CapMatch achieves values of higher than 85.00% on these datasets, outperforming 14 semi-supervised algorithms. When the proportion of labeled data reaches 30%, CapMatch obtains values of no lower than 88.00% on the datasets above, which is better than several classical supervised algorithms, e.g., decision tree and -nearest neighbor (KNN).
Author	Xiao, Zhiwen Song, Fuhong Tong, Huagang Feng, Li Zhu, Zonghai Qu, Rong Luo, Shouxi Xing, Huanlai
Author_xml	– sequence: 1 givenname: Zhiwen orcidid: 0000-0001-9651-111X surname: Xiao fullname: Xiao, Zhiwen email: xiao1994zw@163.com organization: School of Computing and Artificial Intelligence, Southwest Jiaotong University, Chengdu, China – sequence: 2 givenname: Huagang surname: Tong fullname: Tong, Huagang email: huagangtong@gmail.com organization: College of Economic and Management, Nanjing Tech University, Nanjing, China – sequence: 3 givenname: Rong orcidid: 0000-0001-8318-7509 surname: Qu fullname: Qu, Rong email: rong.qu@nottingham.ac.uk organization: School of Computer Science, University of Nottingham, Nottingham, U.K – sequence: 4 givenname: Huanlai orcidid: 0000-0002-6345-7265 surname: Xing fullname: Xing, Huanlai email: hxx@home.swjtu.edu.cn organization: School of Computing and Artificial Intelligence, Southwest Jiaotong University, Chengdu, China – sequence: 5 givenname: Shouxi orcidid: 0000-0002-4041-3681 surname: Luo fullname: Luo, Shouxi email: sxluo@swjtu.edu.cn organization: School of Computing and Artificial Intelligence, Southwest Jiaotong University, Chengdu, China – sequence: 6 givenname: Zonghai orcidid: 0000-0002-2915-0964 surname: Zhu fullname: Zhu, Zonghai email: zzhu@swjtu.edu.cn organization: School of Computing and Artificial Intelligence, Southwest Jiaotong University, Chengdu, China – sequence: 7 givenname: Fuhong orcidid: 0009-0007-1482-3744 surname: Song fullname: Song, Fuhong email: fhsong@mail.gufe.edu.cn organization: School of Information, Guizhou University of Finance and Economics, Guiyang, China – sequence: 8 givenname: Li surname: Feng fullname: Feng, Li email: fengli@swjtu.edu.cn organization: School of Computing and Artificial Intelligence, Southwest Jiaotong University, Chengdu, China
BackLink	https://www.ncbi.nlm.nih.gov/pubmed/38150344$$D View this record in MEDLINE/PubMed
BookMark	eNp9kU1v1DAQhi1UREvpH0AI-dhLtv7Ih8OtbCmt2BaJXQS3aOJMWqPEXmynqFd-Od7utkIc8GUs-X1mPO_7kuxZZ5GQ15zNOGf1yer6erGcCSbkTMo8F3X-jBwIXopMSKX2nu7V931yFMIPlk7JijKvX5B9qXjBEnVAfs9hfQVR376jSxxNtpzW6O9MwI7OnY0eQjR3SFcebOidH9HTRIRpQPrNxFt6jhAnj9l72CCfrPs1YHeD9MwkcBggGmdpAunFNIKlpzq1M_GefkHtbqzZPL8iz3sYAh7t6iH5ev5hNb_IFp8_Xs5PF5mWpYpZp6DHSkDZqpZBC9C1pRK9qHTdVVLJQvd5BW3asO-A5yAYQCF4XQIve647eUiOt33X3v2cMMRmNEFj-qRFN4VG1KzidVHlRZK-3UmndsSuWXszgr9vHn1LArEVaO9C8Ng_SThrNvk0D_k0m3yaXT4JUv9A2sQHh5LPZvg_-maLGkT8a5asCs6F_AMWsqBu
CODEN	ITNNAL
CitedBy_id	crossref_primary_10_1016_j_eswa_2024_124146 crossref_primary_10_1007_s12530_024_09629_x crossref_primary_10_1109_OJCS_2025_3565185 crossref_primary_10_1016_j_engappai_2024_108987 crossref_primary_10_1007_s13198_024_02521_6 crossref_primary_10_1007_s40747_024_01615_9 crossref_primary_10_1007_s40747_024_01643_5 crossref_primary_10_1007_s11042_024_20505_3 crossref_primary_10_1016_j_bspc_2024_107148 crossref_primary_10_1016_j_knosys_2024_112120 crossref_primary_10_1007_s10489_024_06090_w crossref_primary_10_1016_j_engappai_2024_109320 crossref_primary_10_1016_j_neunet_2024_106978 crossref_primary_10_1007_s10489_024_06113_6 crossref_primary_10_1007_s10489_024_05982_1 crossref_primary_10_1007_s10845_025_02632_2 crossref_primary_10_1007_s11042_025_21086_5 crossref_primary_10_1109_TNNLS_2025_3556317 crossref_primary_10_1016_j_bspc_2024_106769 crossref_primary_10_1016_j_ins_2024_121057 crossref_primary_10_1007_s13198_024_02699_9 crossref_primary_10_1007_s13198_025_02803_7 crossref_primary_10_1177_09544070241272761 crossref_primary_10_1007_s11042_024_19895_1 crossref_primary_10_1016_j_bspc_2024_107199 crossref_primary_10_1007_s12530_024_09579_4 crossref_primary_10_1016_j_bspc_2024_106581 crossref_primary_10_1016_j_amc_2023_128412 crossref_primary_10_1177_18761364241296439 crossref_primary_10_1007_s40815_025_01996_0 crossref_primary_10_1007_s40747_024_01441_z crossref_primary_10_1007_s40747_024_01650_6 crossref_primary_10_1016_j_eswa_2024_125946 crossref_primary_10_1007_s40747_024_01461_9 crossref_primary_10_1016_j_ins_2025_122099 crossref_primary_10_1016_j_neunet_2024_106700 crossref_primary_10_1007_s13198_025_02754_z crossref_primary_10_1109_TIM_2025_3550246 crossref_primary_10_1007_s12530_024_09620_6 crossref_primary_10_1016_j_engappai_2024_109236 crossref_primary_10_1007_s10489_024_05819_x crossref_primary_10_1007_s11517_025_03382_2 crossref_primary_10_1016_j_engappai_2024_108586 crossref_primary_10_1016_j_neucom_2024_128115 crossref_primary_10_1007_s11517_024_03095_y crossref_primary_10_1016_j_eswa_2025_128236 crossref_primary_10_1016_j_ins_2024_121742 crossref_primary_10_1007_s10489_024_06025_5 crossref_primary_10_1016_j_engappai_2025_111801 crossref_primary_10_1016_j_knosys_2024_111729 crossref_primary_10_1109_TIM_2025_3563045 crossref_primary_10_1007_s11517_024_03076_1 crossref_primary_10_1016_j_engappai_2024_108921 crossref_primary_10_1016_j_bspc_2024_107008 crossref_primary_10_1016_j_bspc_2024_106553 crossref_primary_10_1016_j_engappai_2024_108966 crossref_primary_10_1007_s13198_024_02568_5 crossref_primary_10_1007_s13721_025_00513_5 crossref_primary_10_1007_s10489_024_05964_3 crossref_primary_10_1007_s13198_024_02480_y crossref_primary_10_1016_j_engappai_2024_108653 crossref_primary_10_1016_j_engappai_2024_109069 crossref_primary_10_1016_j_jfranklin_2024_107143 crossref_primary_10_1016_j_neucom_2025_130830 crossref_primary_10_3390_ai6090218 crossref_primary_10_1007_s11042_024_20075_4 crossref_primary_10_1109_LSP_2025_3569477 crossref_primary_10_1007_s12530_024_09639_9 crossref_primary_10_1016_j_eswa_2024_124554 crossref_primary_10_1016_j_bspc_2024_106128 crossref_primary_10_1016_j_bspc_2024_106568 crossref_primary_10_1016_j_engappai_2024_107940 crossref_primary_10_1016_j_engappai_2024_109800 crossref_primary_10_1007_s11042_024_20379_5 crossref_primary_10_1016_j_engappai_2024_108437 crossref_primary_10_1016_j_bspc_2024_107336 crossref_primary_10_1016_j_engappai_2024_109011 crossref_primary_10_1007_s11517_024_03227_4 crossref_primary_10_1016_j_engappai_2024_109652 crossref_primary_10_1007_s40815_025_01998_y crossref_primary_10_1007_s10489_024_05811_5 crossref_primary_10_1007_s40815_025_01976_4 crossref_primary_10_1007_s13198_024_02472_y crossref_primary_10_1016_j_eswa_2024_125681 crossref_primary_10_1016_j_ins_2024_120712 crossref_primary_10_1016_j_knosys_2024_112154
Cites_doi	10.1109/TNNLS.2021.3068344 10.1109/TCSS.2020.3014128 10.1109/TIM.2022.3201203 10.1016/j.knosys.2021.106934 10.1109/CVPR46437.2021.01057 10.1145/1964897.1964918 10.1109/IJCB52358.2021.9484410 10.1609/aaai.v33i01.33017699 10.1007/s10994-019-05855-6 10.1109/CVPR42600.2020.00975 10.1109/TCYB.2021.3125320 10.1109/TMM.2022.3156938 10.1109/TII.2013.2255061 10.1109/TPAMI.2009.83 10.1109/TIM.2021.3111996 10.1109/TASLP.2021.3105013 10.1109/TCYB.2021.3090370 10.1007/978-3-642-35395-6_30 10.1109/TNNLS.2023.3271140 10.1145/3448112 10.1109/TGRS.2023.3241342 10.1016/j.neucom.2015.07.085 10.1016/j.patcog.2018.12.026 10.3390/s20195707 10.1109/TII.2020.2976812 10.1109/BSN.2016.7516235 10.1016/j.eswa.2019.04.057 10.1109/JIOT.2019.2949715 10.1109/JBHI.2020.3023246 10.1109/TPAMI.2020.2992393 10.48550/arXiv.1503.02531 10.1145/2487575.2487633 10.1109/TNNLS.2021.3053576 10.1109/TIP.2022.3148814 10.1016/j.pmcj.2014.05.006 10.1109/TII.2022.3209672 10.1109/TIM.2022.3164145 10.1109/TNNLS.2020.2978942 10.1145/2971648.2971701 10.1109/ISWC.2008.4911590 10.1109/TIP.2021.3101158 10.1109/TIM.2022.3158427 10.1109/TNNLS.2019.2927224 10.1109/TSMCA.2010.2093883 10.1109/JIOT.2018.2823084 10.1109/LGRS.2021.3069799 10.1109/TNNLS.2023.3288139 10.1038/nature14539 10.1109/TGRS.2021.3089929 10.1145/3517246 10.1109/TIM.2020.2968161 10.1007/s00521-021-05793-2 10.1109/RTCSA.2007.17 10.1109/TIM.2022.3164162 10.1109/CVPR.2019.00302 10.1007/978-3-030-58555-6_18 10.1109/TPAMI.2013.50 10.1109/JBHI.2019.2918412 10.1109/TMM.2022.3192663 10.1007/s11263-021-01453-z 10.1007/s10115-017-1090-9 10.1109/TII.2017.2712746 10.1109/TII.2022.3165875 10.48550/ARXIV.1706.03762 10.1109/ISWC.2009.24 10.1007/s12652-022-03768-2 10.1155/2022/1391906 10.1016/j.neucom.2019.12.150 10.1109/JSYST.2022.3153503 10.1109/TCYB.2020.3007506 10.1016/j.knosys.2021.107338 10.1109/TIM.2021.3102735
ContentType	Journal Article
DBID	97E RIA RIE AAYXX CITATION CGR CUY CVF ECM EIF NPM 7X8
DOI	10.1109/TNNLS.2023.3344294
DatabaseName	IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE Electronic Library (IEL) CrossRef Medline MEDLINE MEDLINE (Ovid) MEDLINE MEDLINE PubMed MEDLINE - Academic
DatabaseTitle	CrossRef MEDLINE Medline Complete MEDLINE with Full Text PubMed MEDLINE (Ovid) MEDLINE - Academic
DatabaseTitleList	MEDLINE MEDLINE - Academic
Database_xml	– sequence: 1 dbid: NPM name: PubMed url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed sourceTypes: Index Database – sequence: 2 dbid: RIE name: IEEE Electronic Library (IEL) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher – sequence: 3 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database
DeliveryMethod	fulltext_linktorsrc
Discipline	Computer Science
EISSN	2162-2388
EndPage	2704
ExternalDocumentID	38150344 10_1109_TNNLS_2023_3344294 10375112
Genre	orig-research Research Support, Non-U.S. Gov't Journal Article
GrantInformation_xml	– fundername: Humanities and Social Sciences Program of the Ministry of Education grantid: 23YJCZH201 – fundername: Fundamental Research Funds for the Central Universities, China funderid: 10.13039/501100012226 – fundername: Natural Science Foundation of Hebei Province grantid: F2022105027 funderid: 10.13039/501100003787 – fundername: Natural Science Foundation of Sichuan Province grantid: 2022NSFSC0568; 2022NSFSC0944; 2023NSFSC0459 funderid: 10.13039/501100018542
GroupedDBID	0R~ 4.4 5VS 6IK 97E AAJGR AARMG AASAJ AAWTH ABAZT ABQJQ ABVLG ACIWK ACPRK AENEX AFRAH AGQYO AGSQL AHBIQ AKJIK AKQYR ALMA_UNASSIGNED_HOLDINGS ATWAV BEFXN BFFAM BGNUA BKEBE BPEOZ EBS EJD IFIPE IPLJI JAVBF M43 MS~ O9- OCL PQQKQ RIA RIE RNS AAYXX CITATION CGR CUY CVF ECM EIF NPM RIG 7X8
ID	FETCH-LOGICAL-c368t-d8afe72a6b8b0abaadb682f27c9d73835cf47ab564fda14a20aa52196a16f1cd3
IEDL.DBID	RIE
ISICitedReferencesCount	119
ISICitedReferencesURI	http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001134412500001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN	2162-237X 2162-2388
IngestDate	Thu Oct 02 12:55:07 EDT 2025 Mon Jul 21 05:55:31 EDT 2025 Sat Nov 29 01:40:29 EST 2025 Tue Nov 18 22:18:28 EST 2025 Wed Aug 27 01:52:59 EDT 2025
IsDoiOpenAccess	false
IsOpenAccess	true
IsPeerReviewed	false
IsScholarly	true
Issue	2
Language	English
License	https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html https://doi.org/10.15223/policy-029 https://doi.org/10.15223/policy-037
LinkModel	DirectLink
MergedId	FETCHMERGED-LOGICAL-c368t-d8afe72a6b8b0abaadb682f27c9d73835cf47ab564fda14a20aa52196a16f1cd3
Notes	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23
ORCID	0000-0002-2915-0964 0009-0007-1482-3744 0000-0002-4041-3681 0000-0001-9651-111X 0000-0001-8318-7509 0000-0002-6345-7265
OpenAccessLink	https://nottingham-repository.worktribe.com/output/29001505
PMID	38150344
PQID	2907195745
PQPubID	23479
PageCount	15
ParticipantIDs	proquest_miscellaneous_2907195745 pubmed_primary_38150344 crossref_citationtrail_10_1109_TNNLS_2023_3344294 ieee_primary_10375112 crossref_primary_10_1109_TNNLS_2023_3344294
PublicationCentury	2000
PublicationDate	2025-02-01
PublicationDateYYYYMMDD	2025-02-01
PublicationDate_xml	– month: 02 year: 2025 text: 2025-02-01 day: 01
PublicationDecade	2020
PublicationPlace	United States
PublicationPlace_xml	– name: United States
PublicationTitle	IEEE transaction on neural networks and learning systems
PublicationTitleAbbrev	TNNLS
PublicationTitleAlternate	IEEE Trans Neural Netw Learn Syst
PublicationYear	2025
Publisher	IEEE
Publisher_xml	– name: IEEE
References	ref13 ref57 ref12 ref56 ref15 ref59 ref14 ref58 ref52 ref11 Tang (ref46) 2020 ref55 ref10 ref54 Liu (ref72) ref17 ref16 ref19 ref18 Berthelot (ref27) Xie (ref84) ref51 ref50 ref45 ref48 ref47 ref42 ref41 ref85 ref44 ref43 Zhang (ref30) ref49 ref8 ref7 ref9 ref4 ref3 ref6 ref5 ref82 ref81 ref40 Duan (ref74) 2020 ref83 Sohn (ref29) Sabour (ref34) ref80 Berthelot (ref28) ref35 ref79 ref78 ref37 ref36 ref31 ref33 ref77 Anguita (ref1) ref32 ref76 ref2 ref39 ref38 Romero (ref66) ref71 ref70 ref73 DeVries (ref75) 2017 ref24 ref68 ref23 ref67 ref26 ref25 Chen (ref53) ref69 ref20 ref64 ref63 ref22 ref21 ref65 ref60 ref62 ref61
References_xml	– ident: ref60 doi: 10.1109/TNNLS.2021.3068344 – ident: ref73 doi: 10.1109/TCSS.2020.3014128 – ident: ref8 doi: 10.1109/TIM.2022.3201203 – start-page: 3857 volume-title: Proc. Adv. Neural Inf. Proces. Syst. ident: ref34 article-title: Dynamic routing between capsules – ident: ref37 doi: 10.1016/j.knosys.2021.106934 – ident: ref76 doi: 10.1109/CVPR46437.2021.01057 – ident: ref78 doi: 10.1145/1964897.1964918 – ident: ref25 doi: 10.1109/IJCB52358.2021.9484410 – ident: ref23 doi: 10.1609/aaai.v33i01.33017699 – ident: ref15 doi: 10.1007/s10994-019-05855-6 – ident: ref52 doi: 10.1109/CVPR42600.2020.00975 – ident: ref64 doi: 10.1109/TCYB.2021.3125320 – ident: ref36 doi: 10.1109/TMM.2022.3156938 – ident: ref40 doi: 10.1109/TII.2013.2255061 – ident: ref39 doi: 10.1109/TPAMI.2009.83 – ident: ref13 doi: 10.1109/TIM.2021.3111996 – ident: ref59 doi: 10.1109/TASLP.2021.3105013 – ident: ref55 doi: 10.1109/TCYB.2021.3090370 – ident: ref79 doi: 10.1007/978-3-642-35395-6_30 – ident: ref54 doi: 10.1109/TNNLS.2023.3271140 – ident: ref24 doi: 10.1145/3448112 – ident: ref67 doi: 10.1109/TGRS.2023.3241342 – ident: ref77 doi: 10.1016/j.neucom.2015.07.085 – ident: ref81 doi: 10.1016/j.patcog.2018.12.026 – ident: ref49 doi: 10.3390/s20195707 – ident: ref45 doi: 10.1109/TII.2020.2976812 – ident: ref42 doi: 10.1109/BSN.2016.7516235 – ident: ref7 doi: 10.1016/j.eswa.2019.04.057 – ident: ref43 doi: 10.1109/JIOT.2019.2949715 – ident: ref57 doi: 10.1109/JBHI.2020.3023246 – start-page: 437 volume-title: Proc. 21st Eur. Symp. Artif. Neural Netw., Comput. Intell. Mach. Learn. ident: ref1 article-title: A public domain dataset for human activity recognition using smartphones – year: 2017 ident: ref75 article-title: Improved regularization of convolutional neural networks with cutout publication-title: arXiv:1708.04552 – ident: ref33 doi: 10.1109/TPAMI.2020.2992393 – ident: ref62 doi: 10.48550/arXiv.1503.02531 – ident: ref82 doi: 10.1145/2487575.2487633 – ident: ref2 doi: 10.1109/TNNLS.2021.3053576 – ident: ref56 doi: 10.1109/TIP.2022.3148814 – start-page: 1 volume-title: Proc. Adv. Neural Inf. Proces. Syst. ident: ref84 article-title: Unsupervised data augmentation for consistency training – ident: ref85 doi: 10.1016/j.pmcj.2014.05.006 – ident: ref70 doi: 10.1109/TII.2022.3209672 – ident: ref4 doi: 10.1109/TIM.2022.3164145 – ident: ref14 doi: 10.1109/TNNLS.2020.2978942 – ident: ref17 doi: 10.1145/2971648.2971701 – ident: ref21 doi: 10.1109/ISWC.2008.4911590 – ident: ref63 doi: 10.1109/TIP.2021.3101158 – start-page: 1 volume-title: Proc. Adv. Neural Inf. Proces. Syst. ident: ref30 article-title: FlexMatch: Boosting semi-supervised learning with curriculum pseudo labeling – ident: ref47 doi: 10.1109/TIM.2022.3158427 – start-page: 66 volume-title: Proc. FinNLP@IJCAI ident: ref72 article-title: Transformer-based capsule network for stock movements prediction – ident: ref19 doi: 10.1109/TNNLS.2019.2927224 – ident: ref9 doi: 10.1109/TSMCA.2010.2093883 – ident: ref44 doi: 10.1109/JIOT.2018.2823084 – ident: ref58 doi: 10.1109/LGRS.2021.3069799 – ident: ref61 doi: 10.1109/TNNLS.2023.3288139 – ident: ref11 doi: 10.1038/nature14539 – ident: ref50 doi: 10.1109/TGRS.2021.3089929 – ident: ref26 doi: 10.1145/3517246 – ident: ref35 doi: 10.1109/TIM.2020.2968161 – start-page: 1 volume-title: Proc. Adv. Neural Inf. Proces. Syst. ident: ref29 article-title: FixMatch: Simplifying semi-supervised learning with consistency and confidence – ident: ref83 doi: 10.1007/s00521-021-05793-2 – ident: ref20 doi: 10.1109/RTCSA.2007.17 – ident: ref3 doi: 10.1109/TIM.2022.3164162 – ident: ref69 doi: 10.1109/CVPR.2019.00302 – ident: ref71 doi: 10.1007/978-3-030-58555-6_18 – ident: ref31 doi: 10.1109/TPAMI.2013.50 – start-page: 1575 volume-title: Proc. Int. Conf. Machin. Learn. ident: ref53 article-title: A simple framework for contrastive learning of visual representations – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref28 article-title: RemixMatch: Semi-supervised learning with distribution alignment and augmentation – ident: ref41 doi: 10.1109/JBHI.2019.2918412 – ident: ref68 doi: 10.1109/TMM.2022.3192663 – ident: ref32 doi: 10.1007/s11263-021-01453-z – ident: ref80 doi: 10.1007/s10115-017-1090-9 – start-page: 5050 volume-title: Proc. Adv. Neural Inf. Proces. Syst. ident: ref27 article-title: MixMatch: A holistic approach to semi-supervised learning – ident: ref10 doi: 10.1109/TII.2017.2712746 – ident: ref12 doi: 10.1109/TII.2022.3165875 – ident: ref38 doi: 10.48550/ARXIV.1706.03762 – year: 2020 ident: ref46 article-title: Layer-wise training convolutional neural networks with smaller filters for human activity recognition using wearable sensors publication-title: arXiv:2005.03948 – year: 2020 ident: ref74 article-title: Capsule-transformer for neural machine translation publication-title: arXiv:2004.14649 – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref66 article-title: FitNet: Hints for thin deep nets – ident: ref16 doi: 10.1109/ISWC.2009.24 – ident: ref22 doi: 10.1007/s12652-022-03768-2 – ident: ref5 doi: 10.1155/2022/1391906 – ident: ref18 doi: 10.1016/j.neucom.2019.12.150 – ident: ref51 doi: 10.1109/JSYST.2022.3153503 – ident: ref65 doi: 10.1109/TCYB.2020.3007506 – ident: ref6 doi: 10.1016/j.knosys.2021.107338 – ident: ref48 doi: 10.1109/TIM.2021.3102735
SSID	ssj0000605649
Score	2.6940281
Snippet	This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing...
SourceID	proquest pubmed crossref ieee
SourceType	Aggregation Database Index Database Enrichment Source Publisher
StartPage	2690
SubjectTerms	Algorithms Capsule network (CapNet) Classification algorithms contrastive learning (CL) Data mining Feature extraction Human Activities - classification Human activity recognition human activity recognition (HAR) Humans knowledge distillation (KD) Neural Networks, Computer Pattern Recognition, Automated - methods Semantics semi-supervised learning (SSL) similarity learning Supervised Machine Learning Transformers Unsupervised learning Unsupervised Machine Learning Wearable Electronic Devices wearable sensors
Title	CapMatch: Semi-Supervised Contrastive Transformer Capsule With Feature-Based Knowledge Distillation for Human Activity Recognition
URI	https://ieeexplore.ieee.org/document/10375112 https://www.ncbi.nlm.nih.gov/pubmed/38150344 https://www.proquest.com/docview/2907195745
Volume	36
WOSCitedRecordID	wos001134412500001&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: PRVIEE databaseName: IEEE Electronic Library (IEL) customDbUrl: eissn: 2162-2388 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000605649 issn: 2162-237X databaseCode: RIE dateStart: 20120101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE
link	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1Lj9MwELbYFQcuLI8FymNlJG7I3cR52OG2LKyQgArRInqLxi9tpZJWTcMP4Jcz4yQVHBaJWw6ZKNI3nvnG9szH2CuAgEQUvPAYH0WOOU8AaCOy3DgfSot1ro5iE2o208tl9WVoVo-9MN77ePnMT-kxnuW7je1oq-ycetqIIByxI6XKvlnrsKGSIDEvI92VaSmFzNRybJJJqvPFbPZpPiWt8GmW5RiESZAHs1VBI-_-yklRZOVmvhnzztXJf_7xPXZ3IJj8oveI--yWbx6wk1G8gQ9r-SH7dQnbzxiHr9_wuf-xEvNuS2Gj9Y7TxKodtBQI-WIktmiLFm239vz7an_NiTt2Oy_eApl8HLfm-DsKGuv-hh1HQx6PCfiF7WUq-NfxxtKmOWXfrt4vLj-IQZBB2KzUe-E0BK8klEabBAyAM6WWQSpbOYWlbmFDrsAgDMFBmoNMAJAeVCWkZUityx6x42bT-CeMpyFT0oZCGYk1GhUtJtNW2Vz5xHmTT1g6QlLbYVo5iWas61i1JFUdEa0J0XpAdMJeH2y2_ayOf759Snj98WYP1YS9HKGvcanR-Qk0ftO1tayQj1WFyosJe9z7xMF6dKWnN3z1GbsjSTk43vd-zo73u86_YLftz_2q3Z2hPy_1WfTn3xgW8Vo
linkProvider	IEEE
linkToHtml	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1Nb9QwELWgINEL5aPAQgEjcUPeJo4TJ9xKoSrqNkLsIvYWjb_UlbbZ1WbDD-gvx-MkKzgUiVsOmSjSG8-8sT3zCHkP4DwRBcusj49M-JzHAHLFEqGMdZn2dW4exCZkWebzefGtb1YPvTDW2nD5zI7xMZzlm5VucavsGHvakCDcJfdSIXjUtWvttlQiT82zQHh5nHHGEzkf2mSi4nhWlpPpGNXCx0nirQuU5PH5KsWhd39lpSCzcjvjDJnn7OA___kRedhTTHrS-cRjcsfWT8jBIN9A-9X8lNycwvrSR-Krj3Rqrxds2q4xcDTWUJxZtYEGQyGdDdTW23qLpl1a-nOxvaLIHtuNZZ8ATS6GzTn6GcPGsrtjR70hDQcF9ER3QhX0-3BnaVUfkh9nX2an56yXZGA6yfItMzk4KzlkKlcRKACjspw7LnVhpC92U-2EBOVhcAZiATwC8AShyCDOXKxN8ozs1avaviA0donk2qVScV-lYdmiklxLLaSNjFViROIBkkr388pRNmNZhbolKqqAaIWIVj2iI_JhZ7PupnX88-1DxOuPNzuoRuTdAH3lFxueoEBtV21T8cIzsiKVIh2R551P7KwHV3p5y1ffkgfns8tJNflaXrwi-xx1hMPt7yOyt9209jW5r39tF83mTfDq3xge87k
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=CapMatch%3A+Semi-Supervised+Contrastive+Transformer+Capsule+With+Feature-Based+Knowledge+Distillation+for+Human+Activity+Recognition&rft.jtitle=IEEE+transaction+on+neural+networks+and+learning+systems&rft.au=Xiao%2C+Zhiwen&rft.au=Tong%2C+Huagang&rft.au=Qu%2C+Rong&rft.au=Xing%2C+Huanlai&rft.date=2025-02-01&rft.pub=IEEE&rft.issn=2162-237X&rft.volume=36&rft.issue=2&rft.spage=2690&rft.epage=2704&rft_id=info:doi/10.1109%2FTNNLS.2023.3344294&rft_id=info%3Apmid%2F38150344&rft.externalDocID=10375112
thumbnail_l	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2162-237X&client=summon
thumbnail_m	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2162-237X&client=summon
thumbnail_s	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2162-237X&client=summon