Home > Research > Publications & Outputs > Discounted multi-armed bandit problems on a col...
View graph of relations

Discounted multi-armed bandit problems on a collection of machines with varying speeds

Research output: Contribution to journalJournal article

Published
<mark>Journal publication date</mark>2004
<mark>Journal</mark>Mathematics of Operations Research
Issue number2
Volume29
Number of pages14
Pages (from-to)266-279
Publication statusPublished
Original languageEnglish

Abstract

This paper is the first to consider general multiarmed bandit problems on parallel machines working at different speeds. Block allocation policies make a once-for-all allocation of bandits to machines at time zero. In this class we describe how to achieve Blackwell optimality under given conditions. The block allocation policy identified allocates the bandits with the largest guaranteed reward rates to the machines operating at greatest speed. This policy is shown to be average-reward optimal in the class of general (nonanticipative, nonidling) policies.

Bibliographic note

RAE_import_type : Journal article RAE_uoa_type : Statistics and Operational Research