A Novel Sine Step Size for Warm-Restart Stochastic Gradient Descent

This paper proposes a novel sine step size for warm-restart stochastic gradient descent (SGD). For the SGD based on the new proposed step size, we establish convergence rates for smooth non-convex functions with and without the Polyak–Łojasiewicz (PL) condition. To assess the effectiveness of the ne...

Full description

Saved in:
Bibliographic Details
Main Authors: Mahsa Soheil Shamaee, Sajad Fathi Hafshejani
Format: Article
Language:English
Published: MDPI AG 2024-12-01
Series:Axioms
Subjects:
Online Access:https://www.mdpi.com/2075-1680/13/12/857
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850086476616826880
author Mahsa Soheil Shamaee
Sajad Fathi Hafshejani
author_facet Mahsa Soheil Shamaee
Sajad Fathi Hafshejani
author_sort Mahsa Soheil Shamaee
collection DOAJ
description This paper proposes a novel sine step size for warm-restart stochastic gradient descent (SGD). For the SGD based on the new proposed step size, we establish convergence rates for smooth non-convex functions with and without the Polyak–Łojasiewicz (PL) condition. To assess the effectiveness of the new step size, we implemented it across several datasets, including FashionMNIST, CIFAR10, and CIFAR100. This implementation was compared against eight distinct existing methods. The experimental results demonstrate that the proposed sine step size improves the test accuracy of the CIFAR100 dataset by <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>1.14</mn><mo>%</mo></mrow></semantics></math></inline-formula>. This improvement highlights the efficiency of the new step size when compared to eight other popular step size methods.
format Article
id doaj-art-37e25f9b4aeb48639181b09ebbc0cb47
institution DOAJ
issn 2075-1680
language English
publishDate 2024-12-01
publisher MDPI AG
record_format Article
series Axioms
spelling doaj-art-37e25f9b4aeb48639181b09ebbc0cb472025-08-20T02:43:28ZengMDPI AGAxioms2075-16802024-12-01131285710.3390/axioms13120857A Novel Sine Step Size for Warm-Restart Stochastic Gradient DescentMahsa Soheil Shamaee0Sajad Fathi Hafshejani1Department of Computer Science, Faculty of Mathematical Science, University of Kashan, Kashan 8731753153, IranDepartment of Math and Computer Science, University of Lethbridge, Lethbridge, AB T1K 3M4, CanadaThis paper proposes a novel sine step size for warm-restart stochastic gradient descent (SGD). For the SGD based on the new proposed step size, we establish convergence rates for smooth non-convex functions with and without the Polyak–Łojasiewicz (PL) condition. To assess the effectiveness of the new step size, we implemented it across several datasets, including FashionMNIST, CIFAR10, and CIFAR100. This implementation was compared against eight distinct existing methods. The experimental results demonstrate that the proposed sine step size improves the test accuracy of the CIFAR100 dataset by <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>1.14</mn><mo>%</mo></mrow></semantics></math></inline-formula>. This improvement highlights the efficiency of the new step size when compared to eight other popular step size methods.https://www.mdpi.com/2075-1680/13/12/857stochastic gradient descentdecay step sizeconvergence rate
spellingShingle Mahsa Soheil Shamaee
Sajad Fathi Hafshejani
A Novel Sine Step Size for Warm-Restart Stochastic Gradient Descent
Axioms
stochastic gradient descent
decay step size
convergence rate
title A Novel Sine Step Size for Warm-Restart Stochastic Gradient Descent
title_full A Novel Sine Step Size for Warm-Restart Stochastic Gradient Descent
title_fullStr A Novel Sine Step Size for Warm-Restart Stochastic Gradient Descent
title_full_unstemmed A Novel Sine Step Size for Warm-Restart Stochastic Gradient Descent
title_short A Novel Sine Step Size for Warm-Restart Stochastic Gradient Descent
title_sort novel sine step size for warm restart stochastic gradient descent
topic stochastic gradient descent
decay step size
convergence rate
url https://www.mdpi.com/2075-1680/13/12/857
work_keys_str_mv AT mahsasoheilshamaee anovelsinestepsizeforwarmrestartstochasticgradientdescent
AT sajadfathihafshejani anovelsinestepsizeforwarmrestartstochasticgradientdescent
AT mahsasoheilshamaee novelsinestepsizeforwarmrestartstochasticgradientdescent
AT sajadfathihafshejani novelsinestepsizeforwarmrestartstochasticgradientdescent