FiBeR adds a closed-form filter-aware correction A(ω)σ_w² to the second-moment term for temporally filtered DP gradients, improving adaptive optimization performance.
arXiv preprint arXiv:2011.11660 (2020)
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
years
2026 4representative citing papers
DP-SGD with expected or batch averaging (EASGM or ASGM) has weaker privacy guarantees than the standard subsampled Gaussian mechanism analysis, confirmed by theoretical re-analysis and audits of libraries including Opacus.
DPSR-CG corrects the privacy accounting for selective release in DPSGD by addressing sampling probability variation and reports strong empirical results on MNIST, CIFAR-10, IMDB, and FMNIST while claiming strict privacy.
Federated learning matches centralized F1 scores for mental health detection from social media but differentially private federated learning drops up to 27 F1 points even at epsilon=50 because noise distorts sparse health and emotion markers.
citing papers explorer
-
FIBER: A Differentially Private Optimizer with Filter-Aware Innovation Bias Correction
FiBeR adds a closed-form filter-aware correction A(ω)σ_w² to the second-moment term for temporally filtered DP gradients, improving adaptive optimization performance.
-
Rethinking the Security of DP-SGD: A Corrected Analysis of Differentially Private Machine Learning
DP-SGD with expected or batch averaging (EASGM or ASGM) has weaker privacy guarantees than the standard subsampled Gaussian mechanism analysis, confirmed by theoretical re-analysis and audits of libraries including Opacus.
-
Revisiting Privacy Amplification by Subsampling in Selective Release DPSGD
DPSR-CG corrects the privacy accounting for selective release in DPSGD by addressing sampling probability variation and reports strong empirical results on MNIST, CIFAR-10, IMDB, and FMNIST while claiming strict privacy.
-
FedMental: Evaluating Federated Learning for Mental Health Detection from Social Media Data
Federated learning matches centralized F1 scores for mental health detection from social media but differentially private federated learning drops up to 27 F1 points even at epsilon=50 because noise distorts sparse health and emotion markers.