ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1906.00670
84
53
v1v2 (latest)

Nonstochastic Multiarmed Bandits with Unrestricted Delays

3 June 2019
Tobias Sommer Thune
Nicolò Cesa-Bianchi
Yevgeny Seldin
ArXiv (abs)PDFHTML
Abstract

We investigate multiarmed bandits with delayed feedback, where the delays need neither be identical nor bounded. We first prove that "delayed" Exp3 achieves the O((KT+D)ln⁡K)O(\sqrt{(KT + D)\ln K} )O((KT+D)lnK​) regret bound conjectured by Cesa-Bianchi et al. [2019] in the case of variable, but bounded delays. Here, KKK is the number of actions and DDD is the total delay over TTT rounds. We then introduce a new algorithm that lifts the requirement of bounded delays by using a wrapper that skips rounds with excessively large delays. The new algorithm maintains the same regret bound, but similar to its predecessor requires prior knowledge of DDD and TTT. For this algorithm we then construct a novel doubling scheme that forgoes the prior knowledge requirement under the assumption that the delays are available at action time (rather than at loss observation time). This assumption is satisfied in a broad range of applications, including interaction with servers and service providers. The resulting oracle regret bound is of order min⁡β(∣Sβ∣+βln⁡K+(KT+Dβ)/β)\min_\beta (|S_\beta|+\beta \ln K + (KT + D_\beta)/\beta)minβ​(∣Sβ​∣+βlnK+(KT+Dβ​)/β), where ∣Sβ∣|S_\beta|∣Sβ​∣ is the number of observations with delay exceeding β\betaβ, and DβD_\betaDβ​ is the total delay of observations with delay below β\betaβ. The bound relaxes to O((KT+D)ln⁡K)O (\sqrt{(KT + D)\ln K} )O((KT+D)lnK​), but we also provide examples where Dβ≪DD_\beta \ll DDβ​≪D and the oracle bound has a polynomially better dependence on the problem parameters.

View on arXiv
Comments on this paper