$(ε, u)$-Adaptive Regret Minimization in Heavy-Tailed Bandits

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Genalti, Gianmarco, Marsigli, Lupo, Gatti, Nicola, Metelli, Alberto Maria
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929240098734080
author Genalti, Gianmarco
Marsigli, Lupo
Gatti, Nicola
Metelli, Alberto Maria
author_facet Genalti, Gianmarco
Marsigli, Lupo
Gatti, Nicola
Metelli, Alberto Maria
contents Heavy-tailed distributions naturally arise in several settings, from finance to telecommunications. While regret minimization under subgaussian or bounded rewards has been widely studied, learning with heavy-tailed distributions only gained popularity over the last decade. In this paper, we consider the setting in which the reward distributions have finite absolute raw moments of maximum order $1+ε$, uniformly bounded by a constant $u<+\infty$, for some $ε\in (0,1]$. In this setting, we study the regret minimization problem when $ε$ and $u$ are unknown to the learner and it has to adapt. First, we show that adaptation comes at a cost and derive two negative results proving that the same regret guarantees of the non-adaptive case cannot be achieved with no further assumptions. Then, we devise and analyze a fully data-driven trimmed mean estimator and propose a novel adaptive regret minimization algorithm, AdaR-UCB, that leverages such an estimator. Finally, we show that AdaR-UCB is the first algorithm that, under a known distributional assumption, enjoys regret guarantees nearly matching those of the non-adaptive heavy-tailed case.
format Preprint
id arxiv_https___arxiv_org_abs_2310_02975
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle $(ε, u)$-Adaptive Regret Minimization in Heavy-Tailed Bandits
Genalti, Gianmarco
Marsigli, Lupo
Gatti, Nicola
Metelli, Alberto Maria
Machine Learning
Artificial Intelligence
Heavy-tailed distributions naturally arise in several settings, from finance to telecommunications. While regret minimization under subgaussian or bounded rewards has been widely studied, learning with heavy-tailed distributions only gained popularity over the last decade. In this paper, we consider the setting in which the reward distributions have finite absolute raw moments of maximum order $1+ε$, uniformly bounded by a constant $u<+\infty$, for some $ε\in (0,1]$. In this setting, we study the regret minimization problem when $ε$ and $u$ are unknown to the learner and it has to adapt. First, we show that adaptation comes at a cost and derive two negative results proving that the same regret guarantees of the non-adaptive case cannot be achieved with no further assumptions. Then, we devise and analyze a fully data-driven trimmed mean estimator and propose a novel adaptive regret minimization algorithm, AdaR-UCB, that leverages such an estimator. Finally, we show that AdaR-UCB is the first algorithm that, under a known distributional assumption, enjoys regret guarantees nearly matching those of the non-adaptive heavy-tailed case.
title $(ε, u)$-Adaptive Regret Minimization in Heavy-Tailed Bandits
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2310.02975