EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Separating hard clean samples from noisy samples with samples’ learning risk for DNN when learning with noisy labels

Lihui Deng, Bo Yang, Zhongfeng Kang, Jiajin Wu, Shaosong Li, Yanping Xiang

Complex & Intelligent Systems · 2024-06 · 10 (3) · 4033-4054

DOI: 10.1007/s40747-024-01360-z ↗

Kurz gefasst

Fachartikel in der Zeitschrift Complex & Intelligent Systems, online erschienen am 01.03.2024. Verfasst von Lihui Deng und 5 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 61 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

L earning with N oisy L abels (LNL) methods aim to improve the accuracy of D eep N eural N etworks (DNNs) when the training set contains samples with noisy or incorrect labels, and have become popular in recent years. Existing popular LNL methods frequently regard samples with high learning difficulty (high-loss and low prediction probability) as noisy samples; however, irregular feature patterns from hard clean samples can also cause high learning difficulty, which can lead to the misclassification of hard clean samples as noisy samples. To address this insufficiency, we propose the S amples’ L earning R isk-based L earning with N oisy L abels (SLRLNL) method. Specifically, we propose to separate noisy samples from hard clean samples using samples’ learning risk, which represents samples’ influence on DNN’s accuracy . We show that samples’ learning risk is comprehensively determined by samples’ learning difficulty as well as samples’ feature similarity to other samples, and thus, compared to existing LNL methods that solely rely on the learning difficulty, our method can better separate hard clean samples from noisy samples, since the former frequently possess irregular feature patterns. Moreover, to extract more useful information from samples with irregular feature patterns (i.e., hard samples), we further propose the R elabeling-based L abel A ugmentation (RLA) process to prevent the memorization of hard noisy samples and better learn the hard clean samples, thus enhancing the learning for hard samples. Empirical studies show that samples’ learning risk can identify noisy samples more accurately, and the RLA process can enhance the learning for hard samples. To evaluate the effectiveness of our method, we compare it with popular existing LNL methods on CIFAR-10, CIFAR-100, Animal-10N, Clothing1M, and Docred. The experimental results indicate that our method outperforms other existing methods. The source code for SLRLNL can be found at https://github.com/yangbo1973/SLRLNL .

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Deng, L., Yang, B., Kang, Z., Wu, J., Li, S., & Xiang, Y. (2024). Separating hard clean samples from noisy samples with samples’ learning risk for DNN when learning with noisy labels. Complex & Intelligent Systems, 10(3), 4033-4054. https://doi.org/10.1007/s40747-024-01360-z

Vancouver

Deng L, Yang B, Kang Z, Wu J, Li S, Xiang Y. Separating hard clean samples from noisy samples with samples’ learning risk for DNN when learning with noisy labels. Complex & Intelligent Systems. 2024;10(3):4033-4054. doi:10.1007/s40747-024-01360-z

BibTeX
@article{Deng2024,
  author = {Deng, Lihui and Yang, Bo and Kang, Zhongfeng and Wu, Jiajin and Li, Shaosong and Xiang, Yanping},
  title = {Separating hard clean samples from noisy samples with samples’ learning risk for DNN when learning with noisy labels},
  journal = {Complex & Intelligent Systems},
  year = {2024},
  volume = {10},
  number = {3},
  pages = {4033--4054},
  doi = {10.1007/s40747-024-01360-z}
}
Kostenlos starten →