Skip to main navigation Skip to search Skip to main content

Exploiting Structure of Uncertainty for Efficient Matroid Semi-Bandits

Research output: Contribution to journalConference articlepeer-review

Abstract

We improve the efficiency of algorithms for stochastic combinatorial semi-bandits. In most interesting problems, state-of-the-art algorithms take advantage of structural properties of rewards, such as independence. However, while being optimal in terms of asymptotic regret, these al-gorithms are inefficient. In our paper, we first reduce their implementation to a specific submod-ular maximization. Then, in case of matroid con-straints, we design adapted approximation rou-tines, thereby providing the first efficient algo-rithms that rely on reward structure to improve regret bound. In particular, we improve the state-of-the-art efficient gap-free regret bound by a fac-torm/ log m, where m is the maximum action size. Finally, we show how our improvement translates to more general budgeted combinato-rial semi-bandits.

Original languageEnglish
Pages (from-to)5123-5132
Number of pages10
JournalProceedings of Machine Learning Research
Volume97
Publication statusPublished - 1 Jan 2019
Externally publishedYes
Event36th International Conference on Machine Learning, ICML 2019 - Long Beach, United States
Duration: 9 Jun 201915 Jun 2019

Fingerprint

Dive into the research topics of 'Exploiting Structure of Uncertainty for Efficient Matroid Semi-Bandits'. Together they form a unique fingerprint.

Cite this