AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Renyi Differential Privacy of Propose-Test-Release and Applications to Private and Robust Machine Learning

Tianhao Wang, Saeed Mahloujifar, Shouda Wang, Ruoxi Jia, Prateek Mittal

NeurIPS 2022 · Federated learning and privacy › Differential privacy · approach: Private data analysis frameworks · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use PTR instead of Gaussian noise scaled to global sensitivity for private trimmed-mean SGDTSGD · MNIST, all reported corruption settings and privacy budgetsaccuracy: 87.53% -> 91.43%; 81.31% -> 92.8%Improved
Use PTR instead of Gaussian noise scaled to global sensitivity for private trimmed-mean SGDTSGD · CIFAR10, all reported corruption settings and privacy budgetsaccuracy: 57.94% -> 58.92%; 58.82% -> 60.13%Improved
Use PTR instead of Gaussian noise scaled to global sensitivity for private trimmed-mean SGDTSGD · EMNIST, all reported corruption settings and privacy budgetsaccuracy: 72.94% -> 76.44%; 79.15% -> 81.02%Improved
Use the white-box subsampled PTR RDP bound instead of the black-box general subsampling boundSubsampled PTR privacy analysis · Poisson subsampling; privacy amplification boundRDP bound tightnessImproved
Convert the PTR RDP bound to (ε, δ)-DP instead of directly analyzing (ε, δ)-DPPTR · Privacy guarantee when δ0 is close to target δε privacy parameterImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy