Research

Paper

TESTING March 04, 2026

When to restart? Exploring escalating restarts on convergence

Authors

Ayush K. Varshney, Šarūnas Girdzijauskas, Konstantinos Vandikas, Aneta Vulgarakis Feljan

Abstract

Learning rate scheduling plays a critical role in the optimization of deep neural networks, directly influencing convergence speed, stability, and generalization. While existing schedulers such as cosine annealing, cyclical learning rates, and warm restarts have shown promise, they often rely on fixed or periodic triggers that are agnostic to the training dynamics, such as stagnation or convergence behavior. In this work, we propose a simple yet effective strategy, which we call Stochastic Gradient Descent with Escalating Restarts (SGD-ER). It adaptively increases the learning rate upon convergence. Our method monitors training progress and triggers restarts when stagnation is detected, linearly escalating the learning rate to escape sharp local minima and explore flatter regions of the loss landscape. We evaluate SGD-ER across CIFAR-10, CIFAR-100, and TinyImageNet on a range of architectures including ResNet-18/34/50, VGG-16, and DenseNet-101. Compared to standard schedulers, SGD-ER improves test accuracy by 0.5-4.5%, demonstrating the benefit of convergence-aware escalating restarts for better local optima.

Metadata

arXiv ID: 2603.04117
Provider: ARXIV
Primary Category: cs.LG
Published: 2026-03-04
Fetched: 2026-03-05 06:06

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.04117v1</id>\n    <title>When to restart? Exploring escalating restarts on convergence</title>\n    <updated>2026-03-04T14:35:27Z</updated>\n    <link href='https://arxiv.org/abs/2603.04117v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.04117v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Learning rate scheduling plays a critical role in the optimization of deep neural networks, directly influencing convergence speed, stability, and generalization. While existing schedulers such as cosine annealing, cyclical learning rates, and warm restarts have shown promise, they often rely on fixed or periodic triggers that are agnostic to the training dynamics, such as stagnation or convergence behavior. In this work, we propose a simple yet effective strategy, which we call Stochastic Gradient Descent with Escalating Restarts (SGD-ER). It adaptively increases the learning rate upon convergence. Our method monitors training progress and triggers restarts when stagnation is detected, linearly escalating the learning rate to escape sharp local minima and explore flatter regions of the loss landscape. We evaluate SGD-ER across CIFAR-10, CIFAR-100, and TinyImageNet on a range of architectures including ResNet-18/34/50, VGG-16, and DenseNet-101. Compared to standard schedulers, SGD-ER improves test accuracy by 0.5-4.5%, demonstrating the benefit of convergence-aware escalating restarts for better local optima.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-03-04T14:35:27Z</published>\n    <arxiv:comment>Paper accepted in Sci4DL workshop in ICLR 2026. https://openreview.net/forum?id=18Yf2KKIn0</arxiv:comment>\n    <arxiv:primary_category term='cs.LG'/>\n    <author>\n      <name>Ayush K. Varshney</name>\n    </author>\n    <author>\n      <name>Šarūnas Girdzijauskas</name>\n    </author>\n    <author>\n      <name>Konstantinos Vandikas</name>\n    </author>\n    <author>\n      <name>Aneta Vulgarakis Feljan</name>\n    </author>\n  </entry>"
}