EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

NLOCL: Noise-Labeled Online Continual Learning

Kan Cheng, Yongxin Ma, Guanglu Wang, Linlin Zong, Xinyue Liu

Electronics · 29.06.2024 · 13 (13) · 2560

DOI: 10.3390/electronics13132560 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Electronics, online erschienen am 29.06.2024. Verfasst von Kan Cheng und 4 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 42 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Social Science Planning Foundation of Liaoning Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Continual learning (CL) from infinite data streams has become a challenge for neural network models in real-world scenarios. Catastrophic forgetting of previous knowledge occurs in this learning setting, and existing supervised CL methods rely excessively on accurately labeled samples. However, the real-world data labels are usually misled by noise, which influences the CL agents and aggravates forgetting. To address this problem, we propose a method named noise-labeled online continual learning (NLOCL), which implements the online CL model with noise-labeled data streams. NLOCL uses an empirical replay strategy to retain crucial examples, separates data streams by small-loss criteria, and includes semi-supervised fine-tuning for labeled and unlabeled samples. Besides, NLOCL combines small loss with class diversity measures and eliminates online memory partitioning. Furthermore, we optimized the experience replay stage to enhance the model performance by retaining significant clean-labeled examples and carefully selecting suitable samples. In the experiment, we designed noise-labeled data streams by injecting noisy labels into multiple datasets and partitioning tasks to simulate infinite data streams realistically. The experimental results demonstrate the superior performance and robust learning capabilities of our proposed method.

Förderung

  • Social Science Planning Foundation of Liaoning Province

Zitieren

APA

Cheng, K., Ma, Y., Wang, G., Zong, L., & Liu, X. (2024). NLOCL: Noise-Labeled Online Continual Learning. Electronics, 13(13), 2560. https://doi.org/10.3390/electronics13132560

Vancouver

Cheng K, Ma Y, Wang G, Zong L, Liu X. NLOCL: Noise-Labeled Online Continual Learning. Electronics. 2024;13(13):2560. doi:10.3390/electronics13132560

BibTeX
@article{Cheng2024,
  author = {Cheng, Kan and Ma, Yongxin and Wang, Guanglu and Zong, Linlin and Liu, Xinyue},
  title = {NLOCL: Noise-Labeled Online Continual Learning},
  journal = {Electronics},
  year = {2024},
  volume = {13},
  number = {13},
  pages = {2560},
  doi = {10.3390/electronics13132560}
}
Kostenlos starten →