Research map / Federated learning and privacy / Differential privacy
Differentially Private Statistical Inference through $\beta$-Divergence One Posterior Sampling
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 beta-divergence Bayes one-posterior sampling instead of other DP logistic regression methods as n grows | DP logistic regression · Simulated data | RMSE to true parameter | Helped |
| Increase privacy budget epsilon | betaD-Bayes advantage in logistic regression · Simulated data | RMSE | Helped |
| Use betaD-Bayes for logistic regression on UCI data | DP logistic regression · UCI (adult, abalone and others) | ROC-AUC | Helped |
| Fixed regulariser lambda=1/9 instead of lambda=1/(9n) in the objective-perturbation baseline (ablation) | DP logistic regression baseline · Simulated data | RMSE as epsilon increases | Hurt |
| Use betaD-Bayes instead of DPSGD for small n | one-layer NN regression · Simulated and UCI data | test RMSE | Hurt |
| Use betaD-Bayes instead of DPSGD for large n | one-layer NN regression · Simulated and UCI data | test RMSE | Helped |
| Increase epsilon | all DP logistic regression methods · Membership inference, n=1000, d=2 | attack success rate | Hurt |
| Use betaD-Bayes at fixed attack success rate | DP logistic regression · Membership inference, n=1000, d=2 | RMSE | Helped |
Closest papers by meaning
- On Inherent Privacy of Posterior Sampling: A Unified R\'enyi-Divergence Framework (NeurIPS 2026 · similarity 0.94)
- Privacy Guarantees in Posterior Sampling under Contamination (NeurIPS 2026 · similarity 0.92)
- Differentially Private Bayesian Linear Regression (NeurIPS 2019 · similarity 0.92)
- Exactly Minimax-Optimal Locally Differentially Private Sampling (NeurIPS 2024 · similarity 0.88)
- Private Statistical Estimation via Truncation (NeurIPS 2025 · similarity 0.88)
- Private Gradient Descent for Linear Regression: Tighter Error Bounds and Instance-Specific Uncertainty Estimation (ICML 2024 · similarity 0.87)
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.