Paper ID: 2205.14839

Adversarial Bandits against Arbitrary Strategies

Jung-hun Kim, Se-Young Yun

We study the adversarial bandit problem against arbitrary strategies, in which $S$ is the parameter for the hardness of the problem and this parameter is not given to the agent. To handle this problem, we adopt the master-base framework using the online mirror descent method (OMD). We first provide a master-base algorithm with simple OMD, achieving $\tilde{O}(S^{1/2}K^{1/3}T^{2/3})$, in which $T^{2/3}$ comes from the variance of loss estimators. To mitigate the impact of the variance, we propose using adaptive learning rates for OMD and achieve $\tilde{O}(\min\{\mathbb{E}[\sqrt{SKT\rho_T(h^\dagger)}],S\sqrt{KT}\})$, where $\rho_T(h^\dagger)$ is a variance term for loss estimators.

Submitted: May 30, 2022