Stabilizing updates in differentially private stochastic gradient descent with buffered rejection

Abstract Differentially private stochastic gradient descent is a standard algorithm for training deep models on sensitive data, but under tight privacy budgets it must add large noise to every step, which slows convergence and reduces accuracy. Selective update methods for differential private stoch...

Полное описание

Библиографические подробности

Главные авторы:	Sifan Deng, Kai Zhang, Weilin Zhang, Huiqin Jiang, Pei-Wei Tsai
Формат:	Artigo
Язык:	Inglês
Опубликовано:	Nature Portfolio 2026-03-01
Серии:	Scientific Reports
Предметы:	Differential privacy Deep learning Buffered rejection mechanism
Online-ссылка:	https://doi.org/10.1038/s41598-026-44009-2
Метки:	Добавить метку Нет меток, Требуется 1-ая метка записи!