Efficient Algorithms for Extreme Bandits - Archive ouverte HAL
Communication Dans Un Congrès Année : 2022

Efficient Algorithms for Extreme Bandits

Résumé

In this paper, we contribute to the Extreme Bandit problem, a variant of Multi-Armed Bandits in which the learner seeks to collect the largest possible reward. We first study the concentration of the maximum of i.i.d random variables under mild assumptions on the tail of the rewards distributions. This analysis motivates the introduction of Quantile of Maxima (QoMax). The properties of QoMax are sufficient to build an Explore-Then-Commit (ETC) strategy, QoMax-ETC, achieving strong asymptotic guarantees despite its simplicity. We then propose and analyze a more adaptive, anytime algorithm, QoMax-SDA, which combines QoMax with a subsampling method recently introduced by Baudry et al. (2021). Both algorithms are more efficient than existing approaches in two aspects (1) they lead to better empirical performance (2) they enjoy a significant reduction of the memory and time complexities.
Fichier principal
Vignette du fichier
RBK22.pdf (7.68 Mo) Télécharger le fichier
Origine Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-03741302 , version 1 (01-08-2022)

Identifiants

Citer

Dorian Baudry, Yoan Russac, Emilie Kaufmann. Efficient Algorithms for Extreme Bandits. International conference on Artificial Intelligence and Statistics (AISTATS), Mar 2022, Virtual Conference, Spain. ⟨hal-03741302⟩
61 Consultations
34 Téléchargements

Altmetric

Partager

More