EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Personalized Hierarchical Federated Learning Algorithm Based on Adaptive Differential Privacy

Huijuan Jia, Jingyu Zhao

Academic Journal of Applied Sciences · 25.03.2026 · 1 (1) · 39-43

DOI: 10.54097/2f97h082 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Academic Journal of Applied Sciences, online erschienen am 25.03.2026. Verfasst von Huijuan Jia und 1 weiteren Autor:innen. Die Arbeit nennt 8 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper proposes ADP-PHFL, an Adaptive Differential Privacy-based Personalized Hierarchical Federated Learning framework. It integrates hierarchical federated learning with personalized training via knowledge transfer, mitigating performance degradation caused by directly applying a global model while reducing communication overhead. An adaptive differential privacy mechanism is introduced to allocate privacy budgets more effectively, avoiding the drawbacks of uniform noise injection and enhancing privacy protection against data leakage. Theoretical analysis establishes the convergence and privacy guarantees of the proposed method. Experimental results demonstrate that ADP-PHFL achieves improved communication efficiency and higher model accuracy.

Zitieren

APA

Jia, H., & Zhao, J. (2026). Personalized Hierarchical Federated Learning Algorithm Based on Adaptive Differential Privacy. Academic Journal of Applied Sciences, 1(1), 39-43. https://doi.org/10.54097/2f97h082

Vancouver

Jia H, Zhao J. Personalized Hierarchical Federated Learning Algorithm Based on Adaptive Differential Privacy. Academic Journal of Applied Sciences. 2026;1(1):39-43. doi:10.54097/2f97h082

BibTeX
@article{Jia2026,
  author = {Jia, Huijuan and Zhao, Jingyu},
  title = {Personalized Hierarchical Federated Learning Algorithm Based on Adaptive Differential Privacy},
  journal = {Academic Journal of Applied Sciences},
  year = {2026},
  volume = {1},
  number = {1},
  pages = {39--43},
  doi = {10.54097/2f97h082}
}
Kostenlos starten →