Research map / Federated learning and privacy / Differential privacy
Differentially Private Learning with Small Public Data
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 PPSGD (public data adjusts privacy budget and clipping, then model reuse) instead of merging public data and training SgdAdv | PPSGD · adult-a, square loss, eps=0.1 | accuracy: 0.6842 -> 0.7941 | Helped |
| PPSGD vs Agd (adaptive budget DP gradient descent, public data merged) | PPSGD · real-world datasets | accuracy / MSE: 0.7863 -> 0.7882 (adult-a hinge eps 0.1); MSE 0.1461 -> 0.1027 (cadata) | Helped |
| PPSGD vs SgdAdv on skin hinge | PPSGD · skin, hinge loss, eps=0.1 | accuracy: 0.9079 -> 0.9069 | No clear effect |
| use public data only (OnlyPub) instead of PPSGD | PPSGD · adult-a, hinge loss, eps=0.1 | accuracy: 0.7882 -> 0.7441 | Hurt |
| vary phi, alpha, beta, lambda parameters (ablation) | PPSGD · Synthetic, 9950 private + 50 public | accuracy | Mixed |
Closest papers by meaning
- Oracle-Efficient Differentially Private Learning with Public Data (NeurIPS 2024 · similarity 0.91)
- Public-data Assisted Private Stochastic Optimization: Power and Limitations (NeurIPS 2024 · similarity 0.90)
- Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement (NeurIPS 2024 · similarity 0.90)
- Optimal Differentially Private Model Training with Public Data (ICML 2024 · similarity 0.90)
- Certification for Differentially Private Prediction in Gradient-Based Training (ICML 2025 · similarity 0.89)
- The importance of feature preprocessing for differentially private linear optimization (ICLR 2024 · similarity 0.89)
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.