EUVIMEDEUVIMED entdecken
Crossref

Balancing Feature Alignment and Uniformity for Few-Shot Classification

Yunlong Yu, Dingyi Zhang, Zhong Ji, Xi Li, Jungong Han, Zhongfei Zhang

IEEE Transactions on Image Processing · 2025 · 34 · 4456-4469

DOI: 10.1109/tip.2023.3328475 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Image Processing, 2025. Verfasst von Yunlong Yu und 5 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref) und nennt 77 Literaturangaben. Gefördert von: Key Research and Development Program of Zhejiang Province, China, Zhejiang Provincial Natural Science Foundation of China, National Science Fund for Distinguished Young Scholars.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In Few-Shot Learning (FSL), the objective is to correctly recognize new samples from novel classes with only a few available samples per class. Existing methods in FSL primarily focus on learning transferable knowledge from base classes by maximizing the information between feature representations and their corresponding labels. However, this approach may suffer from the "supervision collapse" issue, which arises due to a bias towards the base classes. In this paper, we propose a solution to address this issue by preserving the intrinsic structure of the data and enabling the learning of a generalized model for the novel classes. Following the InfoMax principle, our approach maximizes two types of mutual information (MI): between the samples and their feature representations, and between the feature representations and their class labels. This allows us to strike a balance between discrimination (capturing class-specific information) and generalization (capturing common characteristics across different classes) in the feature representations. To achieve this, we adopt a unified framework that perturbs the feature embedding space using two low-bias estimators. The first estimator maximizes the MI between a pair of intra-class samples, while the second estimator maximizes the MI between a sample and its augmented views. This framework effectively combines knowledge distillation between class-wise pairs and enlarges the diversity in feature representations. By conducting extensive experiments on popular FSL benchmarks, our proposed approach achieves comparable performances with state-of-the-art competitors. For example, we achieved an accuracy of 69.53% on the miniImageNet dataset and 77.06% on the CIFAR-FS dataset for the 5-way 1-shot task.

Förderung

  • Key Research and Development Program of Zhejiang Province, China
  • Zhejiang Provincial Natural Science Foundation of China
  • National Science Fund for Distinguished Young Scholars
  • Aeronautical Science Foundation of China
  • Science and Technology Innovation 2025 Major Project of Ningbo

Zitieren

APA

Yu, Y., Zhang, D., Ji, Z., Li, X., Han, J., & Zhang, Z. (2025). Balancing Feature Alignment and Uniformity for Few-Shot Classification. IEEE Transactions on Image Processing, 34, 4456-4469. https://doi.org/10.1109/tip.2023.3328475

Vancouver

Yu Y, Zhang D, Ji Z, Li X, Han J, Zhang Z. Balancing Feature Alignment and Uniformity for Few-Shot Classification. IEEE Transactions on Image Processing. 2025;34:4456-4469. doi:10.1109/tip.2023.3328475

BibTeX
@article{Yu2025,
  author = {Yu, Yunlong and Zhang, Dingyi and Ji, Zhong and Li, Xi and Han, Jungong and Zhang, Zhongfei},
  title = {Balancing Feature Alignment and Uniformity for Few-Shot Classification},
  journal = {IEEE Transactions on Image Processing},
  year = {2025},
  volume = {34},
  pages = {4456--4469},
  doi = {10.1109/tip.2023.3328475}
}
Kostenlos starten →