Skip to yearly menu bar Skip to main content


Poster

Nonparametric Contextual Bandits in Metric Spaces with Unknown Metric

Nirandika Wanigasekara · Christina Yu

East Exhibition Hall B + C #7

Keywords: [ Collaborative Filtering ] [ Algorithms -> Clustering; Algorithms ] [ Algorithms ] [ Bandit Algorithms ]


Abstract: Consider a nonparametric contextual multi-arm bandit problem where each arm $a \in [K]$ is associated to a nonparametric reward function $f_a: [0,1] \to \mathbb{R}$ mapping from contexts to the expected reward. Suppose that there is a large set of arms, yet there is a simple but unknown structure amongst the arm reward functions, e.g. finite types or smooth with respect to an unknown metric space. We present a novel algorithm which learns data-driven similarities amongst the arms, in order to implement adaptive partitioning of the context-arm space for more efficient learning. We provide regret bounds along with simulations that highlight the algorithm's dependence on the local geometry of the reward functions.

Live content is unavailable. Log in and register to view live content