Finite Sample Differentially Private Confidence Intervals
read the original abstract
We study the problem of estimating finite sample confidence intervals of the mean of a normal population under the constraint of differential privacy. We consider both the known and unknown variance cases and construct differentially private algorithms to estimate confidence intervals. Crucially, our algorithms guarantee a finite sample coverage, as opposed to an asymptotic coverage. Unlike most previous differentially private algorithms, we do not require the domain of the samples to be bounded. We also prove lower bounds on the expected size of any differentially private confidence set showing that our the parameters are optimal up to polylogarithmic factors.
This paper has not been read by Pith yet.
Forward citations
Cited by 3 Pith papers
-
Private Rate-Double-Robust Inference
Local privacy mechanisms preserve rate-double-robustness, enabling unbiased and semiparametrically efficient inference on target parameters indexed linearly by infinite-dimensional and nonlinearly by low-dimensional c...
-
Exact Unlearning in Reinforcement Learning
For any ρ>0 there exists a ρ-TV-stable RL algorithm for tabular MDPs supporting exact unlearning at expected cost ρ√(ln T) of retraining from scratch, with regret O(H²√(SAT)+H³S²A+H^{2.5}S²A/ρ) and matching lower boun...
-
Differentially private hypothesis testing in survival analysis
Initiates finite-sample theory for differentially private hypothesis testing in survival analysis, with private tests for Cox models and cumulative hazards plus minimax bounds.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.