Feather: An Elegant Solution to Effective DNN Sparsification

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Georgoulakis, Athanasios Glentis, Retsinas, George, Maragos, Petros
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915108301570048
author Georgoulakis, Athanasios Glentis
Retsinas, George
Maragos, Petros
author_facet Georgoulakis, Athanasios Glentis
Retsinas, George
Maragos, Petros
contents Neural Network pruning is an increasingly popular way for producing compact and efficient models, suitable for resource-limited environments, while preserving high performance. While the pruning can be performed using a multi-cycle training and fine-tuning process, the recent trend is to encompass the sparsification process during the standard course of training. To this end, we introduce Feather, an efficient sparse training module utilizing the powerful Straight-Through Estimator as its core, coupled with a new thresholding operator and a gradient scaling technique, enabling robust, out-of-the-box sparsification performance. Feather's effectiveness and adaptability is demonstrated using various architectures on the CIFAR dataset, while on ImageNet it achieves state-of-the-art Top-1 validation accuracy using the ResNet-50 architecture, surpassing existing methods, including more complex and computationally heavy ones, by a considerable margin. Code is publicly available at https://github.com/athglentis/feather .
format Preprint
id arxiv_https___arxiv_org_abs_2310_02448
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Feather: An Elegant Solution to Effective DNN Sparsification
Georgoulakis, Athanasios Glentis
Retsinas, George
Maragos, Petros
Machine Learning
Neural Network pruning is an increasingly popular way for producing compact and efficient models, suitable for resource-limited environments, while preserving high performance. While the pruning can be performed using a multi-cycle training and fine-tuning process, the recent trend is to encompass the sparsification process during the standard course of training. To this end, we introduce Feather, an efficient sparse training module utilizing the powerful Straight-Through Estimator as its core, coupled with a new thresholding operator and a gradient scaling technique, enabling robust, out-of-the-box sparsification performance. Feather's effectiveness and adaptability is demonstrated using various architectures on the CIFAR dataset, while on ImageNet it achieves state-of-the-art Top-1 validation accuracy using the ResNet-50 architecture, surpassing existing methods, including more complex and computationally heavy ones, by a considerable margin. Code is publicly available at https://github.com/athglentis/feather .
title Feather: An Elegant Solution to Effective DNN Sparsification
topic Machine Learning
url https://arxiv.org/abs/2310.02448