Research map / Federated learning and privacy / Differential privacy
LAPRAS : Learning-Augmented PRivate Answering for linear query Streams.
Pranay Mundra, Adam Sealfon, Ziteng Sun, Quanquan C. Liu
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 |
|---|---|---|---|
| increase overlap of predictions from 0.0 to 1.0 with matrix_heavy split | LAPRAS Static · Adult (S=100, eps=1.0) | median MAE: 368.5622 -> 14.3340 | Improved |
| increase overlap of predictions from 0.0 to 1.0 with matrix_heavy split | LAPRAS Static · Gowalla (S=100, eps=1.0) | median MAE: 391.1077 -> 17.0740 | Improved |
| use LAPRAS query_heavy instead of Independent Noise at zero overlap | LAPRAS Static · Adult (overlap 0.0) | median MAE: 186.5037 -> 201.7689 | Hurt |
| use LAPRAS query_heavy instead of Independent Noise at zero overlap | LAPRAS Static · Gowalla (overlap 0.0) | median MAE: 204.0514 -> 213.8990 | Hurt |
| add Smooth Allocation instead of Static Allocation, matrix_heavy, overlap 0.7 (ablation) | LAPRAS · Adult (|P|=50, Q=50) | MAE: 137.3 -> 84.5 | Improved |
| add Smooth Allocation instead of Static Allocation at zero overlap (ablation) | LAPRAS query_heavy · Adult (|P|=100, S=100) | MAE: 194.6 -> 325.7 | Hurt |
| add query cache (Smooth+Cache) at overlap 0.8 matrix_heavy (ablation) | LAPRAS Smooth · Adult (|P|=100, Q=50) | MAE: 64.1 -> 41.8 | Improved |
| add query cache (Smooth+Cache) at overlap 0.5 query_heavy (ablation) | LAPRAS Smooth · Adult (|P|=100, Q=50) | MAE: 107.0 -> 110.5 | Hurt |
| evaluate under bad-first adversarial arrival order with Smooth instead of Static, query_heavy (ablation) | LAPRAS · Gowalla (overlap 0.5) | MAE: 130.8 -> 162.4 | Hurt |
| replace CacheDP with LAPRAS Static | CacheDP · Adult (overlap 0.0) | NMAE: 0.2076 -> 0.0019 | Improved |
| increase predicted set size from |P|=50 to |P|=100 (more false positives) (ablation) | LAPRAS Smooth · Adult (overlap 1.0, S=50) | Smooth MAE: 74.6 -> 72.4 | No clear effect |
| use matrix_heavy instead of query_heavy budget split at overlap 1.0 (ablation) | LAPRAS Static · Adult (S=100) | median MAE: 43.0020 -> 14.3340 | Improved |
Closest papers by meaning
- Private Prediction via Shrinkage (NeurIPS 2026 · similarity 0.89)
- Efficient and Private Marginal Reconstruction with Local Non-Negativity (NeurIPS 2024 · similarity 0.88)
- On Differential Privacy for Adaptively Solving Search Problems via Sketching (ICML 2025 · similarity 0.87)
- Scalable DP-SGD: Shuffling vs. Poisson Subsampling (NeurIPS 2024 · similarity 0.87)
- Auditing Private Prediction (ICML 2024 · similarity 0.87)
- The Limits of Differential Privacy in Online Learning (NeurIPS 2024 · 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.
