19
27

Efficient Online Linear Optimization with Approximation Algorithms

Abstract

We revisit the problem of \textit{online linear optimization} in case the set of feasible actions is accessible through an approximated linear optimization oracle with a factor α\alpha multiplicative approximation guarantee. This setting is in particular interesting since it captures natural online extensions of well-studied \textit{offline} linear optimization problems which are NP-hard, yet admit efficient approximation algorithms. The goal here is to minimize the α\alpha\textit{-regret} which is the natural extension of the standard \textit{regret} in \textit{online learning} to this setting. We present new algorithms with significantly improved oracle complexity for both the full information and bandit variants of the problem. Mainly, for both variants, we present α\alpha-regret bounds of O(T1/3)O(T^{-1/3}), were TT is the number of prediction rounds, using only O(logT)O(\log{T}) calls to the approximation oracle per iteration, on average. These are the first results to obtain both average oracle complexity of O(logT)O(\log{T}) (or even poly-logarithmic in TT) and α\alpha-regret bound O(Tc)O(T^{-c}) for a constant c>0c>0, for both variants.

View on arXiv
Comments on this paper