Versatile Dueling Bandits: Best-of-both-World Analyses for Online Learning from Preferences

TL;DR

This paper introduces a unified algorithm for dueling bandits that performs optimally in both stochastic and adversarial environments, achieving instance-specific regret bounds and robustness to corrupted preferences.

Contribution

It presents the first best-of-both-world algorithm for dueling bandits with optimal regret bounds and a novel reduction from dueling to multi-armed bandits, simplifying analysis and improving guarantees.

Findings

Achieves optimal $O(rac{ ext{log} T}{ ext{gap}_i})$ regret bound against the Condorcet-winner.

Proves robustness and optimal regret in adversarially corrupted preferences.

Empirically outperforms existing dueling bandit algorithms.

Abstract

We study the problem of $K$-armed dueling bandit for both stochastic and adversarial environments, where the goal of the learner is to aggregate information through relative preferences of pair of decisions points queried in an online sequential manner. We first propose a novel reduction from any (general) dueling bandits to multi-armed bandits and despite the simplicity, it allows us to improve many existing results in dueling bandits. In particular, \emph{we give the first best-of-both world result for the dueling bandits regret minimization problem} -- a unified framework that is guaranteed to perform optimally for both stochastic and adversarial preferences simultaneously. Moreover, our algorithm is also the first to achieve an optimal $O(\sum_{i = 1}^K \frac{\log T}{\Delta_i})$ regret bound against the Condorcet-winner benchmark, which scales optimally both in terms of the arm-size $K$ and the instance-specific suboptimality gaps $\{\Delta_i\}_{i = 1}^K$. This resolves the long-standing problem of designing an instancewise gap-dependent order optimal regret algorithm for dueling bandits (with matching lower bounds up to small constant factors). We further justify the robustness of our proposed algorithm by proving its optimal regret rate under adversarially corrupted preferences -- this outperforms the existing state-of-the-art corrupted dueling results by a large margin. In summary, we believe our reduction idea will find a broader scope in solving a diverse class of dueling bandits setting, which are otherwise studied separately from multi-armed bandits with often more complex solutions and worse guarantees. The efficacy of our proposed algorithms is empirically corroborated against the existing dueling bandit methods.