Bounded regret in stochastic multi-armed bandits

We study the stochastic multi-armed bandit problem when one knows the value μ(⋆) of an optimal arm, as a well as a positive lower bound on the smallest positive gap Δ. We propose a new randomized poli