Research map / Federated learning and privacy / Differential privacy
Optimal Unbiased Randomizers for Regression with Label Differential Privacy
Ashwinkumar Badanidiyuru, Badih Ghazi, Pritish Kamath, Ravi Kumar, Ethan Jacob Leeman, Pasin Manurangsi, Avinash V Varadarajan, Chiyuan Zhang
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 the optimal unbiased label randomizer instead of competing randomizers | Regression models trained with label DP · Criteo Sponsored Search Conversion | test mean squared error | Improved |
| Use the optimal unbiased label randomizer instead of competing randomizers | Regression models trained with label DP · US Census | test mean squared error | Improved |
| Use an unbiased randomizer rather than a biased randomizer | Regression models trained with label DP · Criteo Sponsored Search Conversion | test mean squared error | Improved |
| Use an unbiased randomizer rather than a biased randomizer | Regression models trained with label DP · App Ads Conversion Count | relative Poisson loss | Improved |
| Choose the optimal unbiased randomizer rather than Laplace or additive staircase | Regression models trained with label DP · App Ads Conversion Count | relative Poisson loss | Improved |
| Leave Laplace and additive staircase labels unclipped rather than clipping them to the valid label range (ablation) | Laplace and additive staircase mechanisms · Regression datasets with label DP | test performance | Improved |
| Estimate the label prior privately and optimize the unbiased randomizer for that prior | Optimal unbiased label randomizer · Criteo Sponsored Search Conversion | test mean squared error | Improved |
Closest papers by meaning
- Enhancing Learning with Label Differential Privacy by Vector Approximation (ICLR 2025 · similarity 0.89)
- Certification for Differentially Private Prediction in Gradient-Based Training (ICML 2025 · similarity 0.86)
- Differentially Private Analysis for Binary Response Models: Optimality, Estimation, and Inference (ICML 2025 · similarity 0.86)
- Auditing Privacy Mechanisms via Label Inference Attacks (NeurIPS 2024 · similarity 0.85)
- It's My Data Too: Private ML for Datasets with Multi-User Training Examples (ICML 2025 · similarity 0.85)
- Differentially Private Model Merging (NeurIPS 2026 · 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.
