Under central differential privacy, the sharp L1 separation radius for two-sample testing of Hölder-smooth densities is the maximum of the classical rate and three privacy barriers, and adapting to unknown smoothness costs only a log-log factor.
Adaptive goodness-of-fit tests in a density model
1 Pith paper cite this work, alongside 31 external citations. Polarity classification is still indexing.
1
Pith paper citing it
31
external citations · OpenAlex
fields
math.ST 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Sharp Minimax Rates for Smooth Two-Sample Testing under Central Differential Privacy
Under central differential privacy, the sharp L1 separation radius for two-sample testing of Hölder-smooth densities is the maximum of the classical rate and three privacy barriers, and adapting to unknown smoothness costs only a log-log factor.