PulseExploreJournal ClubResearchersJournals
Instagram
HomeJournal ClubExplore
Synapse
⌘+K
Synapse
April 23, 2024EntropyOpen Access

LF3PFL: A Practical Privacy-Preserving Federated Learning Algorithm Based on Local Federalization Scheme

View Full Paper
Ask AI
Bookmark
Share

Authors

YLYong LiGXGaochao XuXMXutao Meng

Discussion

Loading...

Member takes

Overview

Randomized trial demonstrates improved privacy preservation in federated learning, highlighting practical benefits.

Key Points

  • LF3PFL significantly enhances privacy preservation of participant data while maintaining competitive training accuracies in federated learning.
  • The method utilizes localized federated updates and cross-entropy optimization to reduce information loss during training, achieving greater data confidentiality.
  • Experimental validation across CIFAR-10, Shakespeare, and MNIST demonstrates clear effectiveness in training accuracy and privacy protection versus existing methods of federated learning strategies, outperforming standard approaches in real-world scenarios and diverse model assessments with five distinct models used in comparison studies, revealing consistently superior results for localized updates within federated settings as viability is underscored by these empirical results, showing methodological strength and implementation efficiency as central aspects of this research initiative, closing performance gaps and addressing privacy concerns essential for operational federated learning frameworks.

Cite This Study

Li et al. (2024) studied this question.

synapsesocial.com/papers/68e6df83b6db64358765ac5fhttps://doi.org/10.3390/e26050353
View Full Paper
Ask AI
Bookmark
Share