Research map / Federated learning and privacy / Differential privacy
Individual Privacy Accounting via a Rényi Filter
Vitaly Feldman, Tijana Zrnic
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 |
|---|---|---|---|
| Add individual privacy filtering and continue optimization with points whose realized gradient norms leave privacy budget | Private GD · MNIST, tuned regime | test accuracy: 0.3: $(93.29 \pm 0.49)\%$ -> $(93.64 \pm 0.46)\%; 0.5: $(94.62\pm 0.43)\%$ -> $(94.90\pm 0.26)\%; 1.0: $(96.25\pm0.23)\% | Mixed |
| Add individual privacy filtering under suboptimal clipping | Private GD · MNIST, suboptimal clipping | test accuracy: 0.3: $(84.47 \pm 3.95)\%$ -> $(92.25 \pm 0.91)\%; 0.5: $(92.07\pm2.07)\%$ -> $(94.30 \pm 0.58)\%; 1.0: $(94.45 \pm 0.45) | Improved |
| Use private batch gradient descent rather than private stochastic gradient descent | Private MNIST training · MNIST | privacy-utility tradeoff: 94.63% ± 0.34% for ε = 1.16; the paper compares this with ε = 0.5, δ = 10^{-5} | Improved |
Closest papers by meaning
- Auditing Private Prediction (ICML 2024 · similarity 0.87)
- Instance-Specific Asymmetric Sensitivity in Differential Privacy (NeurIPS 2024 · similarity 0.87)
- Privacy Profiles for Private Selection (ICML 2024 · similarity 0.87)
- Auditing $f$-differential privacy in one run (ICML 2025 · similarity 0.86)
- 𝑓-Differential Privacy Filters: Validity and Approximate Solutions (NeurIPS 2026 · similarity 0.86)
- Differentially Private Model Merging (NeurIPS 2026 · similarity 0.86)
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.
