AdaPrivate-TS is a private contextual bandit algorithm that interprets DP noise as increased posterior covariance and uses batched zCDP composition to achieve O(sqrt(d) log T / sqrt(rho)) privacy cost under event-level privacy.
Dynamic Global Sensitivity for Differentially Private Contextual Bandits
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
AdaPrivate-TS: Private Thompson Sampling for Contextual Bandits with Privacy Amplification
AdaPrivate-TS is a private contextual bandit algorithm that interprets DP noise as increased posterior covariance and uses batched zCDP composition to achieve O(sqrt(d) log T / sqrt(rho)) privacy cost under event-level privacy.