Research map / Federated learning and privacy / Differential privacy
CoinPress: Practical Private Mean and Covariance Estimation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| use 3 iterations of CoinPress instead of 1 (naive private estimate) (ablation) | private covariance PCA · POPRES Europe, top PC alignment | inner product with true PC1: 0.48 -> 0.98 | Helped |
| use 5 iterations instead of 1 (ablation) | private covariance PCA · POPRES Europe, top PC alignment | inner product with true PC2: 0.28 -> 0.92 | Helped |
| increase iterations t from 3 to 5 (ablation) | private covariance PCA · POPRES Europe, top PC alignment | inner product with true PC1: 0.98 -> 0.96 (PC1); 0.48 -> 0.92 (PC2) | Mixed |
| use more iterations t>1 instead of t=1 (ablation) | CoinPress covariance estimation · synthetic Gaussian covariance, d=10, rho=0.5 | Mahalanobis error | Helped |
| increase initial radius R by three orders of magnitude, with t=10 (ablation) | CoinPress mean estimation · Gaussian d=50, n=1000, rho=0.5 | l2 error | No clear effect |
| increase initial radius R for baseline methods | prior private mean estimators · Gaussian d=50, n=1000, rho=0.5 | l2 error | Hurt |
Closest papers by meaning
- Private Adaptive Covariance Estimation via Gaussian Graphical Models (NeurIPS 2026 · similarity 0.89)
- Private Statistical Estimation via Truncation (NeurIPS 2025 · similarity 0.87)
- Dimension-free Private Mean Estimation for Anisotropic Distributions (NeurIPS 2024 · similarity 0.86)
- Analysis of Differentially Private Synthetic Data: A Measurement Error Approach (AAAI 2024 · similarity 0.86)
- Better Locally Private Sparse Estimation Given Multiple Samples Per User (ICML 2024 · similarity 0.85)
- On Differentially Private U Statistics (NeurIPS 2024 · similarity 0.85)
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.