Variance-Dependent Regret Bounds for Non-stationary Linear Bandits

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Zhiyong, Xie, Jize, Chen, Yi, Lui, John C. S., Zhou, Dongruo
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914715992588288
author Wang, Zhiyong
Xie, Jize
Chen, Yi
Lui, John C. S.
Zhou, Dongruo
author_facet Wang, Zhiyong
Xie, Jize
Chen, Yi
Lui, John C. S.
Zhou, Dongruo
contents We investigate the non-stationary stochastic linear bandit problem where the reward distribution evolves each round. Existing algorithms characterize the non-stationarity by the total variation budget $B_K$, which is the summation of the change of the consecutive feature vectors of the linear bandits over $K$ rounds. However, such a quantity only measures the non-stationarity with respect to the expectation of the reward distribution, which makes existing algorithms sub-optimal under the general non-stationary distribution setting. In this work, we propose algorithms that utilize the variance of the reward distribution as well as the $B_K$, and show that they can achieve tighter regret upper bounds. Specifically, we introduce two novel algorithms: Restarted Weighted$\text{OFUL}^+$ and Restarted $\text{SAVE}^+$. These algorithms address cases where the variance information of the rewards is known and unknown, respectively. Notably, when the total variance $V_K$ is much smaller than $K$, our algorithms outperform previous state-of-the-art results on non-stationary stochastic linear bandits under different settings. Experimental evaluations further validate the superior performance of our proposed algorithms over existing works.
format Preprint
id arxiv_https___arxiv_org_abs_2403_10732
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Variance-Dependent Regret Bounds for Non-stationary Linear Bandits
Wang, Zhiyong
Xie, Jize
Chen, Yi
Lui, John C. S.
Zhou, Dongruo
Machine Learning
Artificial Intelligence
We investigate the non-stationary stochastic linear bandit problem where the reward distribution evolves each round. Existing algorithms characterize the non-stationarity by the total variation budget $B_K$, which is the summation of the change of the consecutive feature vectors of the linear bandits over $K$ rounds. However, such a quantity only measures the non-stationarity with respect to the expectation of the reward distribution, which makes existing algorithms sub-optimal under the general non-stationary distribution setting. In this work, we propose algorithms that utilize the variance of the reward distribution as well as the $B_K$, and show that they can achieve tighter regret upper bounds. Specifically, we introduce two novel algorithms: Restarted Weighted$\text{OFUL}^+$ and Restarted $\text{SAVE}^+$. These algorithms address cases where the variance information of the rewards is known and unknown, respectively. Notably, when the total variance $V_K$ is much smaller than $K$, our algorithms outperform previous state-of-the-art results on non-stationary stochastic linear bandits under different settings. Experimental evaluations further validate the superior performance of our proposed algorithms over existing works.
title Variance-Dependent Regret Bounds for Non-stationary Linear Bandits
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2403.10732