Asymptotic Optimality for Decentralised Bandits

Conor J. Newton; Ayalvadi Ganesh; Henry W.J. Reeve

doi:10.1007/s13235-022-00451-1

Asymptotic Optimality for Decentralised Bandits

Conor J. Newton^*, Ayalvadi Ganesh, Henry W.J. Reeve

^*Corresponding author for this work

Research output: Contribution to journal › Article (Academic Journal) › peer-review

1 Citation (Scopus)

104 Downloads (Pure)

Abstract

We consider a large number of agents collaborating on a multi-armed bandit problem with a large number of arms. The goal is to minimise the regret of each agent in a communication-constrained setting. We present a decentralised algorithm which builds upon and improves the Gossip-Insert-Eliminate method of Chawla et al. (International conference on artificial intelligence and statistics, pp 3471–3481, 2020). We provide a theoretical analysis of the regret incurred which shows that our algorithm is asymptotically optimal. In fact, our regret guarantee matches the asymptotically optimal rate achievable in the full communication setting. Finally, we present empirical results which support our conclusions.

Original language	English
Pages (from-to)	307-325
Journal	Dynamic Games and Applications
Volume	13
Early online date	20 Jun 2022
DOIs	https://doi.org/10.1007/s13235-022-00451-1
Publication status	Published - 1 Mar 2023

Bibliographical note

Publisher Copyright:
© 2022, The Author(s).

Access to Document

10.1007/s13235-022-00451-1Licence: CC BY

Full-text PDF (final published version)
This is the final published version of the article (version of record). It first appeared online via Springer at https://doi.org/10.1007/s13235-022-00451-1 . Please refer to any applicable terms of use of the publisher.
Final published version, 1.07 MBLicence: CC BY

Handle.net

Persistent link

Cite this

@article{2aa5754611c8458fbab340a1cdf1b776,

title = "Asymptotic Optimality for Decentralised Bandits",

abstract = "We consider a large number of agents collaborating on a multi-armed bandit problem with a large number of arms. The goal is to minimise the regret of each agent in a communication-constrained setting. We present a decentralised algorithm which builds upon and improves the Gossip-Insert-Eliminate method of Chawla et al. (International conference on artificial intelligence and statistics, pp 3471–3481, 2020). We provide a theoretical analysis of the regret incurred which shows that our algorithm is asymptotically optimal. In fact, our regret guarantee matches the asymptotically optimal rate achievable in the full communication setting. Finally, we present empirical results which support our conclusions.",

author = "Newton, \{Conor J.\} and Ayalvadi Ganesh and Reeve, \{Henry W.J.\}",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s).",

year = "2023",

month = mar,

day = "1",

doi = "10.1007/s13235-022-00451-1",

language = "English",

volume = "13",

pages = "307--325",

journal = "Dynamic Games and Applications",

issn = "2153-0785",

publisher = "Springer Science + Business Media",

}

TY - JOUR

T1 - Asymptotic Optimality for Decentralised Bandits

AU - Newton, Conor J.

AU - Ganesh, Ayalvadi

AU - Reeve, Henry W.J.

PY - 2023/3/1

Y1 - 2023/3/1

N2 - We consider a large number of agents collaborating on a multi-armed bandit problem with a large number of arms. The goal is to minimise the regret of each agent in a communication-constrained setting. We present a decentralised algorithm which builds upon and improves the Gossip-Insert-Eliminate method of Chawla et al. (International conference on artificial intelligence and statistics, pp 3471–3481, 2020). We provide a theoretical analysis of the regret incurred which shows that our algorithm is asymptotically optimal. In fact, our regret guarantee matches the asymptotically optimal rate achievable in the full communication setting. Finally, we present empirical results which support our conclusions.

AB - We consider a large number of agents collaborating on a multi-armed bandit problem with a large number of arms. The goal is to minimise the regret of each agent in a communication-constrained setting. We present a decentralised algorithm which builds upon and improves the Gossip-Insert-Eliminate method of Chawla et al. (International conference on artificial intelligence and statistics, pp 3471–3481, 2020). We provide a theoretical analysis of the regret incurred which shows that our algorithm is asymptotically optimal. In fact, our regret guarantee matches the asymptotically optimal rate achievable in the full communication setting. Finally, we present empirical results which support our conclusions.

UR - https://www.scopus.com/pages/publications/85132735272

U2 - 10.1007/s13235-022-00451-1

DO - 10.1007/s13235-022-00451-1

M3 - Article (Academic Journal)

AN - SCOPUS:85132735272

SN - 2153-0785

VL - 13

SP - 307

EP - 325

JO - Dynamic Games and Applications

JF - Dynamic Games and Applications

ER -

Asymptotic Optimality for Decentralised Bandits

Abstract

Bibliographical note

Access to Document

Handle.net

Other files and links

Fingerprint

Asymptotic Optimality for Decentralised Bandits

Cite this