EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Privacy-Enhanced Federated Learning for Non-IID Data

Qingjie Tan, Shuhui Wu, Yuanhong Tao

Mathematics · 29.09.2023 · 11 (19) · 4123

DOI: 10.3390/math11194123 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Mathematics, online erschienen am 29.09.2023. Verfasst von Qingjie Tan und 2 weiteren Autor:innen. Die Arbeit wurde bisher 11-mal zitiert (Crossref) und nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Zhejiang University of Science and Technology.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Federated learning (FL) allows the collaborative training of a collective model by a vast number of decentralized clients while ensuring that these clients’ data remain private and are not shared. In practical situations, the training data utilized in FL often exhibit non-IID characteristics, hence diminishing the efficacy of FL. Our study presents a novel privacy-preserving FL algorithm, HW-DPFL, which leverages data label distribution similarity as a basis for its design. Our proposed approach achieves this objective without incurring any additional overhead communication. In this study, we provide evidence to support the assertion that our approach improves the privacy guarantee and convergence of FL both theoretically and empirically.

Förderung

  • Zhejiang University of Science and Technology

Zitieren

APA

Tan, Q., Wu, S., & Tao, Y. (2023). Privacy-Enhanced Federated Learning for Non-IID Data. Mathematics, 11(19), 4123. https://doi.org/10.3390/math11194123

Vancouver

Tan Q, Wu S, Tao Y. Privacy-Enhanced Federated Learning for Non-IID Data. Mathematics. 2023;11(19):4123. doi:10.3390/math11194123

BibTeX
@article{Tan2023,
  author = {Tan, Qingjie and Wu, Shuhui and Tao, Yuanhong},
  title = {Privacy-Enhanced Federated Learning for Non-IID Data},
  journal = {Mathematics},
  year = {2023},
  volume = {11},
  number = {19},
  pages = {4123},
  doi = {10.3390/math11194123}
}
Kostenlos starten →