On the boundedness of the sequence generated by minibatch stochastic gradient descent

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bauschke, Heinz H., Tung, Tran Thanh
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913917573267456
author Bauschke, Heinz H.
Tung, Tran Thanh
author_facet Bauschke, Heinz H.
Tung, Tran Thanh
contents Stochastic Gradient Descent (SGD) with Polyak's stepsize has recently gained renewed attention in stochastic optimization. Recently, Orvieto, Lacoste-Julien, and Loizou introduced a decreasing variant of Polyak's stepsize, where convergence relies on a boundedness assumption of the iterates. They established that this assumption holds under strong convexity. In this paper, we extend their result by proving that boundedness also holds for a broader class of objective functions, including coercive functions. We also present a case in which boundedness may or may not hold.
format Preprint
id arxiv_https___arxiv_org_abs_2506_23303
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle On the boundedness of the sequence generated by minibatch stochastic gradient descent
Bauschke, Heinz H.
Tung, Tran Thanh
Optimization and Control
Functional Analysis
Primary 90C15, 90C25, 65K05, Secondary 68T07, 68W20, 68W40
Stochastic Gradient Descent (SGD) with Polyak's stepsize has recently gained renewed attention in stochastic optimization. Recently, Orvieto, Lacoste-Julien, and Loizou introduced a decreasing variant of Polyak's stepsize, where convergence relies on a boundedness assumption of the iterates. They established that this assumption holds under strong convexity. In this paper, we extend their result by proving that boundedness also holds for a broader class of objective functions, including coercive functions. We also present a case in which boundedness may or may not hold.
title On the boundedness of the sequence generated by minibatch stochastic gradient descent
topic Optimization and Control
Functional Analysis
Primary 90C15, 90C25, 65K05, Secondary 68T07, 68W20, 68W40
url https://arxiv.org/abs/2506.23303