PP-QADMM: A Dual-Driven Perturbation and Quantized ADMM for Privacy Preserving and Communication-Efficient Federated Learning

This article presents Privacy-Preserving and Quantized ADMM (PP-QADMM), a novel federated learning (FL) algorithm that is both privacy-preserving and communication-efficient, built upon the Alternating Direction Method of Multipliers (ADMM). PP-QADMM enhances privacy through three core mechanisms. F...

Повний опис

Бібліографічні деталі

Автор:	Anis Elgabli
Формат:	Artigo
Мова:	Inglês
Опубліковано:	IEEE 2025-01-01
Серія:	IEEE Open Journal of the Communications Society
Предмети:	Communication-efficient federated learning differential privacy ADMM stochastic quantization secure aggregation
Онлайн доступ:	https://ieeexplore.ieee.org/document/10982181/
Теги:	Додати тег Немає тегів, Будьте першим, хто поставить тег для цього запису!