Research map / Federated learning and privacy / Differential privacy
Discrete Distribution Estimation under Local Privacy
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 projected decoder instead of ML/normalized decoder for k-RR (ablation) | k-RR · geometric distribution simulation | l1 decoding error | Helped |
| use projected decoder instead of standard/normalized decoder for k-Rappor (ablation) | k-Rappor · geometric distribution simulation | l1 decoding error | Helped |
| use normalized decoder instead of projected decoder on a flat Dirichlet distribution (ablation) | k-RR and k-Rappor · Dirichlet distribution simulation | l1 decoding error | Helped |
| k-RR versus k-Rappor when k < e^epsilon | k-RR · simulations over k and epsilon | l1 decoding error | Helped |
| k-RR versus k-Rappor for large k and low epsilon | k-RR · simulations over k and epsilon | l1 decoding error | Hurt |
| increase number of users | k-RR advantage over k-Rappor · low privacy simulations | l1 decoding error gap | Hurt |
| use more cohorts C (C=1 vs larger) (ablation) | O-RR · S=256, n=10^6 geometric | median l1 loss | Helped |
| use more than one cohort (C>=2) (ablation) | O-Rappor · S=256, n=10^6 geometric | median l1 loss | Helped |
| increase C beyond 2 (ablation) | O-Rappor · S=256, n=10^6 geometric | median l1 loss | No clear effect |
| increase hash block size k (ablation) | O-Rappor · S=256, n=10^6 geometric | median l1 loss | Helped |
| use more Bloom filter hash functions h (h>1) (ablation) | O-Rappor · S=256, n=10^6 geometric | median l1 loss | No clear effect |
| replace generic hashes with minimal perfect hash functions for closed alphabets (ablation) | O-RR · closed alphabet simulation | l1 utility | Helped |
Closest papers by meaning
- Learning discrete distributions: user vs item-level privacy (NeurIPS 2020 · similarity 0.89)
- A Generalized Shuffle Framework for Privacy Amplification: Strengthening Privacy Guarantees and Enhancing Utility (AAAI 2024 · similarity 0.88)
- A Huber Loss Minimization Approach to Mean Estimation under User-level Differential Privacy (NeurIPS 2024 · similarity 0.87)
- Lightweight Protocols for Distributed Private Quantile Estimation (ICML 2025 · similarity 0.86)
- Optimal Survey Design for Private Mean Estimation (ICML 2025 · similarity 0.86)
- Universal Exact Compression of Differentially Private Mechanisms (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.