Tractable Instances of Bilinear Maximization: Implementing LinUCB on Ellipsoids

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Raymond, Hadiji, Hédi, Combes, Richard
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912699535851520
author Zhang, Raymond
Hadiji, Hédi
Combes, Richard
author_facet Zhang, Raymond
Hadiji, Hédi
Combes, Richard
contents We consider the maximization of $x^\top θ$ over $(x,θ) \in \mathcal{X} \times Θ$, with $\mathcal{X} \subset \mathbb{R}^d$ convex and $Θ\subset \mathbb{R}^d$ an ellipsoid. This problem is fundamental in linear bandits, as the learner must solve it at every time step using optimistic algorithms. We first show that for some sets $\mathcal{X}$ e.g. $\ell_p$ balls with $p>2$, no efficient algorithms exist unless $\mathcal{P} = \mathcal{NP}$. We then provide two novel algorithms solving this problem efficiently when $\mathcal{X}$ is a centered ellipsoid. Our findings provide the first known method to implement optimistic algorithms for linear bandits in high dimensions.
format Preprint
id arxiv_https___arxiv_org_abs_2511_07504
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Tractable Instances of Bilinear Maximization: Implementing LinUCB on Ellipsoids
Zhang, Raymond
Hadiji, Hédi
Combes, Richard
Machine Learning
We consider the maximization of $x^\top θ$ over $(x,θ) \in \mathcal{X} \times Θ$, with $\mathcal{X} \subset \mathbb{R}^d$ convex and $Θ\subset \mathbb{R}^d$ an ellipsoid. This problem is fundamental in linear bandits, as the learner must solve it at every time step using optimistic algorithms. We first show that for some sets $\mathcal{X}$ e.g. $\ell_p$ balls with $p>2$, no efficient algorithms exist unless $\mathcal{P} = \mathcal{NP}$. We then provide two novel algorithms solving this problem efficiently when $\mathcal{X}$ is a centered ellipsoid. Our findings provide the first known method to implement optimistic algorithms for linear bandits in high dimensions.
title Tractable Instances of Bilinear Maximization: Implementing LinUCB on Ellipsoids
topic Machine Learning
url https://arxiv.org/abs/2511.07504