Risk-Sensitive Online Algorithms
Abstract
We study the design of risk-sensitive online algorithms, in which risk measures are used in the competitive analysis of randomized online algorithms. We introduce the CVaR-competitive ratio (-CR) using the conditional value-at-risk of an algorithm's cost, which measures the expectation of the -fraction of worst outcomes against the offline optimal cost, and use this measure to study three online optimization problems: continuous-time ski rental, discrete-time ski rental, and one-max search. The structure of the optimal -CR and algorithm varies significantly between problems: we prove that the optimal -CR for continuous-time ski rental is , obtained by an algorithm described by a delay differential equation. In contrast, in discrete-time ski rental with buying cost , there is an abrupt phase transition at , after which the classic deterministic strategy is optimal. Similarly, one-max search exhibits a phase transition at , after which the classic deterministic strategy is optimal; we also obtain an algorithm that is asymptotically optimal as that arises as the solution to a delay differential equation.
Cite
@article{arxiv.2405.09859,
title = {Risk-Sensitive Online Algorithms},
author = {Nicolas Christianson and Bo Sun and Steven Low and Adam Wierman},
journal= {arXiv preprint arXiv:2405.09859},
year = {2024}
}
Comments
Accepted for presentation at the Conference on Learning Theory (COLT) 2024. Updated with an additional reference and minor edits