Pushing the Limits of Sparsity:A Bag of Tricks for Extreme Pruning

Li, Andy, Durrant, Aiden ORCID: https://orcid.org/0000-0002-8375-4523, Markovic, Milan, Huang, Tianjin, Kundu, Souvik, Chen, Tianlong, Yin, Lu and Leontidis, Georgios (2025) Pushing the Limits of Sparsity:A Bag of Tricks for Extreme Pruning. Transactions on Machine Learning Research. ISSN 2835-8856

Preview	PDF (5122_Pushing_the_Limits_of_Spa) - Published Version Available under License Unspecified licence. Download (777kB) | Preview
Preview	PDF (2411.13545v4) - Accepted Version Download (923kB) | Preview

Abstract

Pruning of deep neural networks has been an effective technique for reducing model size while preserving most of the performance of dense networks, crucial for deploying models on memory and power-constrained devices. While recent sparse learning methods have shown promising performance up to moderate sparsity levels such as 95% and 98%, accuracy quickly deteriorates when pushing sparsities to extreme levels due to unique challenges such as fragile gradient flow. In this work, we explore network performance beyond the commonly studied sparsities, and develop techniques that encourage stable training without accuracy collapse even at extreme sparsities, including 99.90%, 99.95% and 99.99% on ResNet architectures. We propose three complementary techniques that enhance sparse training through different mechanisms: 1) Dynamic ReLU phasing, where DyReLU initially allows for richer parameter exploration before being gradually replaced by standard ReLU, 2) weight sharing which reuses parameters within a residual layer while maintaining the same number of learnable parameters, and 3) cyclic sparsity, where both sparsity levels and sparsity patterns evolve dynamically throughout training to better encourage parameter exploration. We eval uate our method, which we term Extreme Adaptive Sparse Training (EAST) at extreme sparsities using ResNet-34 and ResNet-50 on CIFAR-10, CIFAR-100, and ImageNet, achieving competitive or improved performance compared to existing methods, with notable gains at extreme sparsity levels. Code is available at https://github.com/TensorStrike/ES2.

Item Type:	Article
Additional Information:	Publisher Copyright: © 2025, Transactions on Machine Learning Research. All rights reserved.
Uncontrolled Keywords:	computer vision and pattern recognition,artificial intelligence ,/dk/atira/pure/subjectarea/asjc/1700/1707
Faculty \ School:	Faculty of Science > School of Computing Sciences
UEA Research Groups:	Faculty of Science > Research Groups > Data Science and AI
Related URLs:	https://www.scopus.com/pages/publication...
Depositing User:	LivePure Connector
Date Deposited:	18 Aug 2026 12:57
Last Modified:	18 Aug 2026 12:57
URI:	https://ueaeprints.uea.ac.uk/id/eprint/104222
DOI:	issn:2835-8856

Downloads

Actions (login required)

View Item