Pure entropic regularization for metrical task systems
Abstract
We show that on every -point HST metric, there is a randomized online algorithm for metrical task systems (MTS) that is -competitive for service costs and -competitive for movement costs. In general, these refined guarantees are optimal up to the implicit constant. While an -competitive algorithm for MTS on HST metrics was developed by Bubeck et al. (SODA 2019), that approach could only establish an -competitive ratio when the service costs are required to be -competitive. Our algorithm can be viewed as an instantiation of online mirror descent with the regularizer derived from a multiscale conditional entropy. In fact, our algorithm satisfies a set of even more refined guarantees; we are able to exploit this property to combine it with known random embedding theorems and obtain, for any -point metric space, a randomized algorithm that is -competitive for service costs and -competitive for movement costs.
Cite
@article{arxiv.1906.04270,
title = {Pure entropic regularization for metrical task systems},
author = {Christian Coester and James R. Lee},
journal= {arXiv preprint arXiv:1906.04270},
year = {2020}
}
Comments
COLT 2019