Olga Aryasova (Inst. of Geophysics, Nat. Acad. of Sciences of Ukraine / Friedrich–Schiller–Univ. Jena)
Dr. Tomáš Kocák
Abstract: In the first part of the talk, we introduce the framework of adversarial bandits, compare it to stochastic bandits, and present the regret analysis for the EXP3 algorithm, that solves the problem. In the second part of the talk, we consider problems with a structure where the learner can receive additional information on top of the traditional bandit feedback.