Online learning in bandits with predicted context

arXiv:2307.13916v3 Announce Type: replace-cross
Abstract: We consider the contextual bandit problem where at each time, the agent only has access to a noisy version of the context and the error variance (or an estimator of this variance). This setting is motivated by a wide range of applications where the true context for decision-making is unobserved, and only a prediction of the context by a potentially complex machine learning algorithm is available. When the context error is non-vanishing, classical bandit algorithms …

abstract agent applications arxiv context cs.lg decision error estimator making online learning stat.ml true type variance