cs.LGJun 22, 2026

Leveraging Similarities in Multi-Armed Bandits

Authors: Khaled EldowaThibaud RahierAugustin CablantPanayotis MertikopoulosPierre Gaillard

Organizations: Univ. Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK, 38000 Grenoble, France · Criteo AI Lab, Paris, France · Univ. Grenoble Alpes, Inria, CNRS, Grenoble INP, LIG, 38000 Grenoble, France

Abstract

In many online learning and bandit problems, the actions we consider possess inherent similarities--for instance because they share latent traits, tags, or hierarchical structure. We study online learning with a similarity-structured action set, encoded by a rooted tree whose leaves are the actions and whose levels quantify how closely two actions are related. The loss sequence is assumed tree-compatible: losses of similar actions are constrained to be close. We establish an impossibility result showing that usual one-point bandit feedback cannot, in general, leverage range or tree-induced similarity, even under very strong similarity constraints. We then provide a unified set of algorithms which adapt to a wide range of richer feedback models, from semi-bandit feedback down to multi-point bandit protocols, including the minimal two-point feedback setting. We show these algorithms exhibit best-of-both-worlds guarantees and provably exploit action similarities by replacing the number of actions KK by a similarity-aware effective number of actions KeffK_{\mathrm{eff}} in the regret bounds. As an application, we show that under two-point feedback, it is possible to achieve T\sqrt{T} regret in Lipschitz bandits when d2d \leq 2.

Explore similar work

CardsList
  1. On-line Learning in Tree MDPs by Treating Policies as Bandit Arms

    May 6, 2026Anvay Shah, Ramsundar Anandanarayanan, Sharayu Moharir +1Markov Decision ProcessesBandits

  2. Multi-Armed Bandits With Best-Action Queries

    May 8, 2026Francesco Bacchiocchi, Matteo Castiglioni, Alberto Marchesi +1Multi-Armed BanditsO(T^Β)$ Simultaneous Regret