Active Learning for Class-Incremental Time Series with Dual Memory and Shapley Replay

https://m2.mtmt.hu/ hun 2026-03-16 16:46 JournalArticle 36524966 ADMIN_APPROVED true This work is funded by the EU KDT-JU organization under grant agreement 101112089, within the project AIMS5.0 and from the partners’ national programs and funding authorities. 0 false 2026-01-23T09:00:50.236+0000 2026-01-23T08:58:00.813+0000 2025-12-19T21:06:49.858+0000 true 10000597 Szűcs Gábor /api/author/10000597 Author Szűcs Gábor (mesterséges intelligencia, adattudomány) true 10000597 2026-01-23T08:58:00.968+0000 2026-01-23T08:58:00.968+0000 2025-12-23T12:35:06.843+0000 true 10072271 Andódy Katalin /api/admin/10072271 Admin Andódy Katalin (BME admin4) true true false true 24 24 /api/publicationtype/24 PublicationType Folyóiratcikk 1 true 24 JournalArticle true 1268360 Conference paper in journal true 24 24 /api/publicationtype/24 PublicationType Folyóiratcikk 1 true 24 JournalArticle /api/subtype/1268360 Konferenciaközlemény SubType Konferenciaközlemény (Folyóiratcikk) 109 true 1268360 true 1 /api/category/1 Category Tudományos true 1 Szűcs, Gábor Active Learning for Class-Incremental Time Series with Dual Memory and Shapley Replay true true /api/journal/10070927 REVIEWED IFAC PAPERSONLINE 2405-8971 2405-8963 true false 10070927 true 10070927 true 2405-8971 2405-8963 Journal FOREIGN 59 26 229 234 229 6 2025 true 351194 /api/conference/351194 true 10722 true 10020 /api/country/10020 Country Olaszország true 16 /api/city/10722 City Padova, Olaszország true 1302361 Conference 7th IFAC Conference on Intelligent Control and Automation Sciences, ICONS 2025 2025-09-15 [Padova, Olaszország] true Class-Incremental Learning (CIL) for time series data addresses the challenge of incrementally learning new classes from sequential data streams without forgetting previously acquired knowledge. In practical scenarios, labeling all incoming data is costly, necessitating efficient strategies for selective labeling. To address this, we introduce a novel framework integrating Active Learning (AL) into CIL specifically for time series classification tasks, enhancing label efficiency under budget constraints. Our proposed method employs a dual-memory architecture comprising an active short-term memory for temporary storage of newly labeled samples and a consolidated long-term memory utilizing a replay mechanism based on Adversarial Shapley value Experience Replay (ASER) guided by approximate Shapley values. At each incremental learning step, our active learning module strategically selects informative and diverse samples for labeling, thereby optimizing the annotation budget. Experiments conducted on benchmark datasets demonstrate significant improvements in accuracy and label efficiency compared to previous methods. true true NATIONAL 2025 true true true true true false GOLD 2026-01-23 true https://doi.org/10.1016/j.ifacol.2025.12.039 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false false false false 2025-12-22T12:35:06.659+0000 Távközlési és Mesterséges Intelligencia Tanszék (BME / VIK) 2 9 1 3 0 false Language 10002 /api/language/10002 Angol Angol English true 2 true PersonAuthorship 129115235 /api/authorship/129115235 Szűcs, Gábor [Szűcs, Gábor (mesterséges intel...), szerző] Távközlési és Mesterséges Intelligencia Tanszék (BME / VIK) 1 0.333 true false false Author 10000597 /api/author/10000597 Szűcs Gábor (mesterséges intelligencia, adattudomány) Szűcs Gábor true 10000597 true Szűcs Gábor true false false AuthorshipType 1 /api/authorshiptype/1 Szerző 0 true 0 true false true PersonAuthorship 129115236 /api/authorship/129115236 Németh, Marcell [Németh, Marcell István (mesterséges intel...), szerző] Távközlési és Mesterséges Intelligencia Tanszék (BME / VIK) 2 0.333 false false false Author 10078864 /api/author/10078864 Németh Marcell István (mesterséges intelligencia, adattudomány) Németh Marcell István true true Németh Marcell true false false AuthorshipType 1 /api/authorshiptype/1 Szerző 0 true 0 true false true PersonAuthorship 129115237 /api/authorship/129115237 Jacsev, Sámuel 3 0.33333334 false true false Jacsev Sámuel true false false AuthorshipType 1 /api/authorshiptype/1 Szerző 0 true 0 true false true PublicationIdentifier 30420452 /api/publicationidentifier/30420452 DOI: 10.1016/j.ifacol.2025.12.039 PlainSource 6 /api/publicationsource/6 DOI PublicationSourceType 10001 /api/publicationsourcetype/10001 DOI true true true DOI DOI https://doi.org/@@@ true true 6 true GOLD true IDENTICAL 10.1016/j.ifacol.2025.12.039 https://doi.org/10.1016/j.ifacol.2025.12.039 false true PublicationIdentifier 31259931 /api/publicationidentifier/31259931 Scopus: 105026927432 PlainSource 3 /api/publicationsource/3 Scopus PublicationSourceType 10003 /api/publicationsourcetype/10003 Indexelő adatbázis false true true Scopus Scopus http://www.scopus.com/record/display.url?origin=inward&eid=2-s2.0-@@@ true true 3 true 105026927432 http://www.scopus.com/record/display.url?origin=inward&eid=2-s2.0-105026927432 false true PublicationIdentifier 30420453 /api/publicationidentifier/30420453 Egyéb URL: https://linkinghub.elsevier.com/retrieve/pii/S2405896325027156 PlainSource 40 /api/publicationsource/40 Egyéb URL PublicationSourceType 10006 /api/publicationsourcetype/10006 Link true true true Egyéb URL Other URL @@@ true true 40 true IDENTICAL https://linkinghub.elsevier.com/retrieve/pii/S2405896325027156 https://linkinghub.elsevier.com/retrieve/pii/S2405896325027156 false true Reference 70935114 /api/reference/70935114 1. Ge 2025: A class-incremental learning method for pcb defect detection 1 false true Reference 70935115 /api/reference/70935115 2. Huang 2024: Class balance matters to active class-incremental learning.., In Proceedings of the 32nd ACM International Conference on Multimedia, p. 9445 2 false true Reference 70935116 /api/reference/70935116 3. Kiyasseh 2021: A clinical deep learning framework for continually learning from cardiac signals across diseases, time, modalities, and institutions.., Nature Communications, 12, p. 4221, DOI: 10.1038/s41467-021-24483-0 3 10.1038/s41467-021-24483-0 false true Reference 70935117 /api/reference/70935117 4. Kwon, Y.D., Chauhan, J., and Mascolo, C. (2021). Fasticarl: Fast incremental classifier and representation learning with efficient budget allocation in audio sensing applications. arXiv preprint arXiv:2106.07268., DOI: 10.21437/Interspeech.2021-1091 4 10.21437/Interspeech.2021-1091 false true Reference 70935118 /api/reference/70935118 5. Li 2017: Learning without forgetting.., IEEE transactions on pattern analysis and machine intelligence, 40, p. 2935, DOI: 10.1109/TPAMI.2017.2773081 5 10.1109/TPAMI.2017.2773081 false true Reference 70935119 /api/reference/70935119 6. Liu 2021: Class-incremental learning for wireless device identification in iot.., IEEE Internet of Things Journal, 8, p. 17227, DOI: 10.1109/JIOT.2021.3078407 6 10.1109/JIOT.2021.3078407 false true Reference 70935120 /api/reference/70935120 7. Pasti 2024: Latent distillation for continual object detection at the edge. In European conference on computer visionp. 279 7 false true Reference 70935121 /api/reference/70935121 8. Qiao 2023: Class-incremental learning on multivariate time series via shape-aligned temporal distillation. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)p. 1 8 false true Reference 70935122 /api/reference/70935122 9. Qiao 2024: Class-incremental learning for time series: Benchmark and evaluation.., In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, p. 5613, DOI: 10.1145/3637528.3671581 9 10.1145/3637528.3671581 false true Reference 70935123 /api/reference/70935123 10. Qiu, X., Hu, J., Zhou, L., Wu, X., Du, J., Zhang, B., Guo, C., Zhou, A., Jensen, C.S., Sheng, Z., et al. (2024). Tfb: Towards comprehensive and fair benchmarking of time series forecasting methods. arXiv preprint arXiv:2403.20150., DOI: 10.14778/3665844.3665863 10 10.14778/3665844.3665863 false true Reference 70935124 /api/reference/70935124 11. Rolnick 2019: Experience replay for continual learning.., Advances in neural information processing systems, 32 11 false true Reference 70935125 /api/reference/70935125 12. Shi 2024: Cross-domain class incremental broad network for continuous diagnosis of rotating machinery faults under variable operating conditions.., IEEE Transactions on Industrial Informatics, 20, p. 6356, DOI: 10.1109/TII.2023.3345449 12 10.1109/TII.2023.3345449 false true Reference 70935126 /api/reference/70935126 13. Shi 2024: Imbalanced class incremental learning system: A task incremental diagnosis method for imbalanced industrial streaming data.., Advanced Engineering Informatics, 62, p. 102832, DOI: 10.1016/j.aei.2024.102832 13 10.1016/j.aei.2024.102832 false true Reference 70935127 /api/reference/70935127 14. Shi 2024: Semi-supervised class incremental broad network for continuous diagnosis of rotating machinery faults with limited labeled samples.., Knowledge-Based Systems, 286, p. 111397, DOI: 10.1016/j.knosys.2024.111397 14 10.1016/j.knosys.2024.111397 false true Reference 70935128 /api/reference/70935128 15. Shim 2021: Online class-incremental continual learning with adversarial shapley value.., In Proceedings of the AAAI Conference on Artificial Intelligence, 35, p. 9630, DOI: 10.1609/aaai.v35i11.17159 15 10.1609/aaai.v35i11.17159 false true Reference 70935129 /api/reference/70935129 16. Sun 2023: Few-shot class-incremental learning for medical time series classification.., IEEE Journal of Biomedical and Health Informatics, 28, p. 1872, DOI: 10.1109/JBHI.2023.3247861 16 10.1109/JBHI.2023.3247861 false true Reference 70935130 /api/reference/70935130 17. Wu, Y., Nie, M., Zhu, T., Chen, L., Ning, H., and Wan, Y. (2025). Ptms-tscil pre-trained models based class-incremental learning. arXiv preprint arXiv:2503.07153. 17 false true Reference 70935131 /api/reference/70935131 18. Yan 2024: Few-shot class-incremental learning for system-level fault diagnosis of wind turbine 18 false true Reference 70935132 /api/reference/70935132 19. Zhang 2024: Time series class-incremental learning with residual mamba encoder. In 2024 5th International Conference on Computers and Artificial Intelligence Technology (CAIT)p. 249 19 false true Reference 70935133 /api/reference/70935133 20. Zhou 2023: Deep class-incremental learning: A survey.., arXiv preprint arXiv:2302.03648, 1, p. 6 20 false true Reference 70935134 /api/reference/70935134 21. Zhou 2024: Class-incremental learning: A survey 21 false true /api/publication/36524966 Szűcs Gábor et al. Active Learning for Class-Incremental Time Series with Dual Memory and Shapley Replay. (2025) IFAC PAPERSONLINE 2405-8971 2405-8963 59 26 229-234<div class="JournalArticle Publication long-list"> <div class="authors"> <img title="Forrásközlemény" style="float: left" src="/frontend/resources/grid/publication-core-icon.png"> <div class="autype autype0"> <span class="author-name" mtid="10000597"><a href="/gui2/?type=authors&mode=browse&sel=10000597" target="_blank">Szűcs Gábor (<span class="authorship-author-name">Szűcs Gábor</span> <span class="authorAux-mtmt"> mesterséges intelligencia, adattudomány</span>) </a> </span> <span class="author-affil"><span title="Budapesti Műszaki és Gazdaságtudományi Egyetem">BME</span>/<span title="Villamosmérnöki és Informatikai Kar">VIK</span>/Távközlési és Mesterséges Intelligencia Tanszék</span> ;    <span class="author-name" mtid="10078864"><a href="/gui2/?type=authors&mode=browse&sel=10078864" target="_blank">Németh Marcell (<span class="authorship-author-name">Németh Marcell István</span> <span class="authorAux-mtmt"> mesterséges intelligencia, adattudomány</span>) </a> </span> <span class="author-affil"><span title="Budapesti Műszaki és Gazdaságtudományi Egyetem">BME</span>/<span title="Villamosmérnöki és Informatikai Kar">VIK</span>/Távközlési és Mesterséges Intelligencia Tanszék</span> ;    <span class="author-name" >Jacsev Sámuel </span> </div> </div> <div class="title"><a href="/gui2/?mode=browse&params=publication;36524966" target="_blank">Active Learning for Class-Incremental Time Series with Dual Memory and Shapley Replay</a></div> <div> <span class="journal-title">IFAC PAPERSONLINE</span> <span class="journal-issn">(<a target="_blank" href="https://portal.issn.org/resource/ISSN/2405-8971">2405-8971</a> <a target="_blank" href="https://portal.issn.org/resource/ISSN/2405-8963">2405-8963</a>)</span>: <span class="journal-volume">59</span> <span class="journal-issue">26</span> <span class="page"> pp 229-234 </span> <span class="year">(2025)</span> <div class="conference"> 7th IFAC Conference on Intelligent Control and Automation Sciences, ICONS 2025. Konferencia helye, ideje: <span class="location">Padova, Olaszország <span class="conference-date">2025.09.15. - 2025.09.18.</span> </div> </div> <div class="pub-footer"> <span class="language" xmlns="http://www.w3.org/1999/html">Nyelv: Angol | </span> <span class="identifiers"> <span class="id identifier oa_GOLD" title=" Gold "> <a style="color:blue" title="10.1016/j.ifacol.2025.12.039" target="_blank" href="https://doi.org/10.1016/j.ifacol.2025.12.039"> DOI </a> </span> <span class="id identifier oa_none" title="none"> <a style="color:black" title="105026927432" target="_blank" href="http://www.scopus.com/record/display.url?origin=inward&eid=2-s2.0-105026927432"> Scopus </a> </span> <span class="id identifier oa_none" title="none"> <a style="color:blue" title="https://linkinghub.elsevier.com/retrieve/pii/S2405896325027156" target="_blank" href="https://linkinghub.elsevier.com/retrieve/pii/S2405896325027156"> Egyéb URL </a> </span> </span> <div class="mtid"><span class="long-pub-mtid">Közlemény: 36524966</span> | <span class="status-data status-ADMIN_APPROVED"> Admin láttamozott </span> Forrás | <span class="type-subtype">Folyóiratcikk ( Konferenciaközlemény ) </span> | <span class="pub-category">Tudományos</span> | <span class="publication-sourceOfData">DOI XML</span> </div> <div class="lastModified">Utolsó módosítás: 2026.01.23. 09:58 Andódy Katalin (BME admin4) </div> <pre class="comment" style="margin-top: 0; margin-bottom: 0;"><u>Megjegyzés</u>: This work is funded by the EU KDT-JU organization under grant agreement 101112089, within the project AIMS5.0 and from the partners’ national programs and funding authorities.</pre> </div></div>