Beating the adaptive bandit with high probability | IEEE Conference Publication | IEEE Xplore