Research map / Federated learning and privacy / Differential privacy
Differentially private methods for managing model uncertainty in linear regression
Víctor Peña, Andrés F. Barrientos
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use Laplace mechanism instead of Wishart mechanism for perturbing the Gram matrix (ablation) | private model averaging · simulation, normal linear model | PMSE | Improved |
| use Wishart mechanism instead of Laplace when p is 6 or 9, true model null and sample small (ablation) | private model averaging · simulation, p=6 or 9, |T|=0, small n | PMSE | Improved |
| increase privacy budget epsilon from 0.5 to 0.9 (ablation) | private methods · simulation, normal linear model | PMSE: epsilon 0.5 -> 0.9 | Improved |
| add hard-thresholding of the perturbed Gram matrix when the true model is null (ablation) | private model averaging · simulation, |T|=0 | PMSE | Improved |
| add hard-thresholding of the perturbed Gram matrix when many predictors are active and n is small (ablation) | private model averaging · simulation, |T| large, small n | PMSE | Hurt |
| increase number of data subsets M in data-splitting hypothesis tests (ablation) | private hypothesis testing · simulation, mean squared error | variance and bias | Mixed |
| use more stringent censoring limits (smaller U-L) in private test statistics (ablation) | private hypothesis testing · simulation and High School and Beyond application | variance and bias | Mixed |
| increase predictors p (ablation) | private model selection · simulation, posterior probability of true model | posterior probability of true model | Hurt |
Closest papers by meaning
- On Inherent Privacy of Posterior Sampling: A Unified R\'enyi-Divergence Framework (NeurIPS 2026 · similarity 0.89)
- On the Computational Complexity of Private High-dimensional Model Selection (NeurIPS 2024 · similarity 0.89)
- Privacy Guarantees in Posterior Sampling under Contamination (NeurIPS 2026 · similarity 0.87)
- Differentially Private Model Merging (NeurIPS 2026 · similarity 0.86)
- Private Gradient Descent for Linear Regression: Tighter Error Bounds and Instance-Specific Uncertainty Estimation (ICML 2024 · similarity 0.86)
- Bounding the Excess Risk for Linear Models Trained on Marginal-Preserving, Differentially-Private, Synthetic Data (ICML 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.
