Salvato in:
Dettagli Bibliografici
Autori principali: Bellavia, Stefania, Krejić, Nataša, Jerinkić, Nataša Krklec, Raydan, Marcos
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:https://arxiv.org/abs/2306.07379
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908078217101312
author Bellavia, Stefania
Krejić, Nataša
Jerinkić, Nataša Krklec
Raydan, Marcos
author_facet Bellavia, Stefania
Krejić, Nataša
Jerinkić, Nataša Krklec
Raydan, Marcos
contents The spectral gradient method is known to be a powerful low-cost tool for solving large-scale optimization problems. In this paper, our goal is to exploit its advantages in the stochastic optimization framework, especially in the case of mini-batch subsampling that is often used in big data settings. To allow the spectral coefficient to properly explore the underlying approximate Hessian spectrum, we keep the same subsample for several iterations before subsampling again. We analyze the required algorithmic features and the conditions for almost sure convergence, and present initial numerical results that show the advantages of the proposed method.
format Preprint
id arxiv_https___arxiv_org_abs_2306_07379
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle SLiSeS: Subsampled Line Search Spectral Gradient Method for Finite Sums
Bellavia, Stefania
Krejić, Nataša
Jerinkić, Nataša Krklec
Raydan, Marcos
Optimization and Control
The spectral gradient method is known to be a powerful low-cost tool for solving large-scale optimization problems. In this paper, our goal is to exploit its advantages in the stochastic optimization framework, especially in the case of mini-batch subsampling that is often used in big data settings. To allow the spectral coefficient to properly explore the underlying approximate Hessian spectrum, we keep the same subsample for several iterations before subsampling again. We analyze the required algorithmic features and the conditions for almost sure convergence, and present initial numerical results that show the advantages of the proposed method.
title SLiSeS: Subsampled Line Search Spectral Gradient Method for Finite Sums
topic Optimization and Control
url https://arxiv.org/abs/2306.07379