1

From optimality to robustness: Dirichlet sampling strategies in stochastic bandits (NeurIPS 2021)

State-of-the-art randomised bandit algorithm with guarantees under weak assumptions.