In the small gap regime, Thompson sampling and a broad class of sampling-based algorithms converge weakly to identical SDE limits, making regret performance insensitive to likelihood misspecification.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2021 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Diffusion Approximations for Thompson Sampling in the Small Gap Regime
In the small gap regime, Thompson sampling and a broad class of sampling-based algorithms converge weakly to identical SDE limits, making regret performance insensitive to likelihood misspecification.