Research map / Federated learning and privacy / Differential privacy
Private Federated Frequency Estimation: Adapting to the Hardness of the Instance
Jingfeng Wu, Wennan Zhu, Peter Kairouz, Vladimir Braverman
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 |
|---|---|---|---|
| set count sketch width by instance-optimal rule instead of minimax-optimal rule (ablation) | count sketch, single round · Gowalla, C4, Sentiment-140 | sketch size for target l_inf error | Improved |
| set count sketch width with the two-phase method instead of minimax-optimal rule (ablation) | count sketch, single round · Gowalla, C4, Sentiment-140 | sketch size for target l_inf error | Improved |
| use two-phase method (small sketch on n/10 clients to fit top-20 polynomial) versus instance-optimal oracle (ablation) | count sketch, single round · Gowalla | sketch size for target l_inf error | No clear effect |
| replace simple count sketch (SCS) with hybrid count sketch (HCS) for heavy-tailed homogeneous frequencies | multi-round FFE · synthetic i^-1.1 and i^-2, d=10^5, M=10 | expected l_inf error | No clear effect |
| use HCS instead of FCS and SCS for steeper power law (degree 3) | multi-round FFE · synthetic homogeneous, M=10 | expected l_inf error | Improved |
| use hybrid or full count sketch instead of simple count sketch with heavy-tailed heterogeneity | multi-round FFE · Gowalla, M=10 | l_inf error | Improved |
| use HCS instead of FCS and SCS under moderately light tails | multi-round FFE · C4, M=10 | estimate shape / error | Improved |
| use HCS in Sentiment-140 multi-round experiment | multi-round FFE · Sentiment-140, M=10 | error | Improved |
| average more rounds in homogeneous case with HCS | hybrid count sketch · homogeneous multi-round FFE | error scaling | Improved |
Closest papers by meaning
- Private and Personalized Frequency Estimation in a Federated Setting (NeurIPS 2024 · similarity 0.89)
- Sketched Gaussian Mechanism for Private Federated Learning (NeurIPS 2025 · similarity 0.85)
- Frequency Oracle for Sensitive Data Monitoring (Student Abstract) (AAAI 2024 · similarity 0.84)
- One-shot Empirical Privacy Estimation for Federated Learning (ICLR 2024 · similarity 0.83)
- Lightweight Protocols for Distributed Private Quantile Estimation (ICML 2025 · similarity 0.83)
- Noise-Aware Algorithm for Heterogeneous Differentially Private Federated Learning (ICML 2024 · similarity 0.83)
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.
