Share on Facebook Tweet on Twitter Share on LinkedIn Share by email
Bandits with concave rewards and convex knapsacks

Shipra Agrawal and Nikhil R. Devanur

Abstract

In this paper, we consider a very general model for exploration-exploitation tradeoff which allows arbitrary concave rewards and convex constraints on the decisions across time, in addition to the customary limitation on the time horizon. This model subsumes the classic multi-armed bandit (MAB) model, and the Bandits with Knapsacks (BwK) model of Badanidiyuru et al.[2013]. We also consider an extension of this model to allow linear contexts, similar to the linear contextual extension of the MAB model. We demonstrate that a natural extension of the UCB family of algorithms for MAB provides a polynomial time algorithm that has near-optimal regret guarantees for this substantially more general model, and matches the bounds provided by Badanidiyuru et al.[2013] for the special case of BwK, which is quite surprising. We also provide computationally more efficient algorithms by establishing interesting connections between this problem and other well studied problems/algorithms such as the Blackwell approachability problem, online convex optimization, and the Frank-Wolfe technique for convex optimization. We give examples of several concrete applications, where this more general model of bandits allows for richer and/or more efficient formulations of the problem.

Details

Publication typeInproceedings
Published inEC 2014
PublisherACM conference on Economics and Computation
> Publications > Bandits with concave rewards and convex knapsacks