For wirelessly aggregated federated learning with smooth non-convex losses, the Rényi-DP privacy loss remains bounded over iterations when the parameter domain is bounded, with a convergence bound for clipped gradients.
Privacy amplification by itera- tion
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
When Differential Privacy Meets Wireless Federated Learning: An Improved Analysis for Privacy and Convergence
For wirelessly aggregated federated learning with smooth non-convex losses, the Rényi-DP privacy loss remains bounded over iterations when the parameter domain is bounded, with a convergence bound for clipped gradients.