Regret Bounds for Batched Bandits

We present simple algorithms for batched stochastic multi-armed bandit and linear problems. prove bounds their expected regrets that improve extend the best known regret of Gao, Han, Ren, Zhou (NeurIPS 2019), any number batches. In particular, our in both settings achieve optimal by using only a logarithmic also study adversarial problem first time provide regret, up to factors, algorithm with predetermined batch sizes.