Optimal Regret with Limited Adaptivity for Generalized Linear Contextual Bandits

arXiv:2404.06831v1 Announce Type: new
Abstract: We study the generalized linear contextual bandit problem within the requirements of limited adaptivity. In this paper, we present two algorithms, \texttt{B-GLinCB} and \texttt{RS-GLinCB}, that address, respectively, two prevalent limited adaptivity models: batch learning with stochastic contexts and rare policy switches with adversarial contexts. For both these models, we establish essentially tight regret bounds. Notably, in the obtained bounds, we manage to eliminate a dependence on a key parameter $\kappa$, which captures the non-linearity of …

abstract adversarial algorithms arxiv cs.lg generalized linear paper policy requirements stochastic study type