Stochastic Gradient Descent with Strategic Querying

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jiang, Nanfei, Wai, Hoi-To, Alizadeh, Mahnoosh
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912551577583616
author Jiang, Nanfei
Wai, Hoi-To
Alizadeh, Mahnoosh
author_facet Jiang, Nanfei
Wai, Hoi-To
Alizadeh, Mahnoosh
contents This paper considers a finite-sum optimization problem under first-order queries and investigates the benefits of strategic querying on stochastic gradient-based methods compared to uniform querying strategy. We first introduce Oracle Gradient Querying (OGQ), an idealized algorithm that selects one user's gradient yielding the largest possible expected improvement (EI) at each step. However, OGQ assumes oracle access to the gradients of all users to make such a selection, which is impractical in real-world scenarios. To address this limitation, we propose Strategic Gradient Querying (SGQ), a practical algorithm that has better transient-state performance than SGD while making only one query per iteration. For smooth objective functions satisfying the Polyak-Lojasiewicz condition, we show that under the assumption of EI heterogeneity, OGQ enhances transient-state performance and reduces steady-state variance, while SGQ improves transient-state performance over SGD. Our numerical experiments validate our theoretical findings.
format Preprint
id arxiv_https___arxiv_org_abs_2508_17144
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Stochastic Gradient Descent with Strategic Querying
Jiang, Nanfei
Wai, Hoi-To
Alizadeh, Mahnoosh
Machine Learning
Optimization and Control
This paper considers a finite-sum optimization problem under first-order queries and investigates the benefits of strategic querying on stochastic gradient-based methods compared to uniform querying strategy. We first introduce Oracle Gradient Querying (OGQ), an idealized algorithm that selects one user's gradient yielding the largest possible expected improvement (EI) at each step. However, OGQ assumes oracle access to the gradients of all users to make such a selection, which is impractical in real-world scenarios. To address this limitation, we propose Strategic Gradient Querying (SGQ), a practical algorithm that has better transient-state performance than SGD while making only one query per iteration. For smooth objective functions satisfying the Polyak-Lojasiewicz condition, we show that under the assumption of EI heterogeneity, OGQ enhances transient-state performance and reduces steady-state variance, while SGQ improves transient-state performance over SGD. Our numerical experiments validate our theoretical findings.
title Stochastic Gradient Descent with Strategic Querying
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2508.17144