Trade-off functions between two distributions are finitely testable if and only if their Neyman-Pearson rejection regions are attainable by a VC-class of sets.
Title resolution pending
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
years
2026 5representative citing papers
Tight closed-form bounds via Berry-Esseen show DP-SGD with random shuffling achieves near-ideal privacy (trade-off close to 1-a) for σ ≥ √(3/ln M) and large M, with δ linear in epochs restricting E to O(√M) and an asymptotic O(√E) δ under E = c_M²M.
Using the shuffle index, the authors formulate and solve an optimization problem for post-shuffle minimax-optimal unbiased mean estimation, yielding an asymptotically optimal mechanism whose privacy-utility tradeoff approaches the central Gaussian mechanism in the high-privacy regime.
A practical federated recommender allows user control over personalization versus diversity, with users showing preference for personalization in a live deployment.
CA-ADP adjusts differential privacy noise per mini-batch class composition to improve F-scores by 3.3-8.5% over standard DP on three fall-detection datasets while claiming formal (ε,δ) guarantees.
citing papers explorer
-
When Are Trade-Off Functions Testable from Finite Samples?
Trade-off functions between two distributions are finitely testable if and only if their Neyman-Pearson rejection regions are attainable by a VC-class of sets.
-
Trade-off Functions for DP-SGD with Subsampling based on Random Shuffling: Tight Upper and Lower Bounds
Tight closed-form bounds via Berry-Esseen show DP-SGD with random shuffling achieves near-ideal privacy (trade-off close to 1-a) for σ ≥ √(3/ln M) and large M, with δ linear in epochs restricting E to O(√M) and an asymptotic O(√E) δ under E = c_M²M.
-
Shuffling-Aware Optimization for Private Vector Mean Estimation
Using the shuffle index, the authors formulate and solve an optimization problem for post-shuffle minimax-optimal unbiased mean estimation, yielding an asymptotically optimal mechanism whose privacy-utility tradeoff approaches the central Gaussian mechanism in the high-privacy regime.
-
Beyond Centralization: User-Controlled Federated Recommendations in Practice
A practical federated recommender allows user control over personalization versus diversity, with users showing preference for personalization in a live deployment.
-
Class-Aware Adaptive Differential Privacy in Deep Learning for Sensor-Based Fall Detection
CA-ADP adjusts differential privacy noise per mini-batch class composition to improve F-scores by 3.3-8.5% over standard DP on three fall-detection datasets while claiming formal (ε,δ) guarantees.