Nonstochastic multi-armed bandits with graph-structured feedback

Noga Alon, Nicolo Cesa-Bianchi, Claudio Gentile, Shie Mannor, Yishay Mansour, Ohad Shamir

Research output: Contribution to journalArticlepeer-review

Abstract

We introduce and study a partial-information model of online learning, where a decision maker repeatedly chooses from a finite set of actions and observes some subset of the associated losses. This setting naturally models several situations where knowing the loss of one action provides information on the loss of other actions. Moreover, it generalizes and interpolates between the well-studied full-information setting (where all losses are revealed) and the bandit setting (where only the loss of the action chosen by the player is revealed). We provide several algorithms addressing different variants of our setting and provide tight regret bounds depending on combinatorial properties of the information feedback structure.

Original languageEnglish
Pages (from-to)1785-1826
Number of pages42
JournalSIAM Journal on Computing
Volume46
Issue number6
DOIs
StatePublished - 2017

Funding

FundersFunder number
FP7/2007
Israel Ministry of Science and Technology
Israeli I-Core program
Oswald Veblen Fund
USA-Israeli BSF
United States - Israel Binational Science Foundation
Seventh Framework Programme306638
United States-Israel Binational Science Foundation
Ministero dell’Istruzione, dell’Università e della Ricerca2010N5K7EB 003
Israel Science Foundation
Seventh Framework Programme
Israeli Centers for Research Excellence425/13

    Keywords

    • Graph theory
    • Learning from experts
    • Learning with partial feedback
    • Multi-armed bandits
    • Online learning

    Fingerprint

    Dive into the research topics of 'Nonstochastic multi-armed bandits with graph-structured feedback'. Together they form a unique fingerprint.

    Cite this