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. 2106.06712
14
4

Simple Combinatorial Algorithms for Combinatorial Bandits: Corruptions and Approximations

12 June 2021
Haike Xu
Jian Li
ArXivPDFHTML
Abstract

We consider the stochastic combinatorial semi-bandit problem with adversarial corruptions. We provide a simple combinatorial algorithm that can achieve a regret of O~(C+d2K/Δmin)\tilde{O}\left(C+d^2K/\Delta_{min}\right)O~(C+d2K/Δmin​) where CCC is the total amount of corruptions, ddd is the maximal number of arms one can play in each round, KKK is the number of arms. If one selects only one arm in each round, we achieves a regret of O~(C+∑Δi>0(1/Δi))\tilde{O}\left(C+\sum_{\Delta_i>0}(1/\Delta_i)\right)O~(C+∑Δi​>0​(1/Δi​)). Our algorithm is combinatorial and improves on the previous combinatorial algorithm by [Gupta et al., COLT2019] (their bound is O~(KC+∑Δi>0(1/Δi))\tilde{O}\left(KC+\sum_{\Delta_i>0}(1/\Delta_i)\right)O~(KC+∑Δi​>0​(1/Δi​))), and almost matches the best known bounds obtained by [Zimmert et al., ICML2019] and [Zimmert and Seldin, AISTATS2019] (up to logarithmic factor). Note that the algorithms in [Zimmert et al., ICML2019] and [Zimmert and Seldin, AISTATS2019] require one to solve complex convex programs while our algorithm is combinatorial, very easy to implement, requires weaker assumptions and has very low oracle complexity and running time. We also study the setting where we only get access to an approximation oracle for the stochastic combinatorial semi-bandit problem. Our algorithm achieves an (approximation) regret bound of O~(dKT)\tilde{O}\left(d\sqrt{KT}\right)O~(dKT​). Our algorithm is very simple, only worse than the best known regret bound by d\sqrt{d}d​, and has much lower oracle complexity than previous work.

View on arXiv
Comments on this paper