Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity

Despite all the recent development and success of deep neural networks, deployment of a deep model onto the resource-constrained devices still remains challenging. However, model pruning can resolve this issue for Convolutional Neural Networks (CNNs), since it is one of the most popular approaches t...

Full description

Saved in:
Bibliographic Details
Main Authors: Niaz Ashraf Khan, A. M. Saadman Rafat
Format: Article
Language:English
Published: Taylor & Francis Group 2025-04-01
Series:Journal of Information and Telecommunication
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/24751839.2024.2415008
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850128829233758208
author Niaz Ashraf Khan
A. M. Saadman Rafat
author_facet Niaz Ashraf Khan
A. M. Saadman Rafat
author_sort Niaz Ashraf Khan
collection DOAJ
description Despite all the recent development and success of deep neural networks, deployment of a deep model onto the resource-constrained devices still remains challenging. However, model pruning can resolve this issue for Convolutional Neural Networks (CNNs), since it is one of the most popular approaches to reducing computational complexities. Therefore, this article presents a pruning model for convolutional neural networks. The proposed method classifies and arranges similar filters into the same cluster where the similarity is calculated using a three-dimensional normalized cross-correlation. Moreover, these steps can be completed entirely based on the filter values while not requiring a set of test images as well as the acquisition of any filter activation. In the research, the performances of the proposed model pruning method have been evaluated, where it is observed that the proposed approach is computationally light and requires significantly less time and resources compared to ML and activation-based approaches. In the experiments, using the VGG16 model on the Cifar10 dataset, the proposed approach results in the pruned model(s) which are comparable in performance with models found using activation-based methods and expensive ML-based methods. Similar results are found when pruning a custom CNN on the MNIST and Fashion MNIST datasets as well.
format Article
id doaj-art-75d70d0a7d4e4e1cb34ea813312b971b
institution OA Journals
issn 2475-1839
2475-1847
language English
publishDate 2025-04-01
publisher Taylor & Francis Group
record_format Article
series Journal of Information and Telecommunication
spelling doaj-art-75d70d0a7d4e4e1cb34ea813312b971b2025-08-20T02:33:11ZengTaylor & Francis GroupJournal of Information and Telecommunication2475-18392475-18472025-04-019219020810.1080/24751839.2024.2415008Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarityNiaz Ashraf Khan0A. M. Saadman Rafat1Department of Computer Science and Engineering, University of Liberal Arts Bangladesh, Dhaka, BangladeshDepartment of Electrical and Computer Engineering, North South University, Dhaka, BangladeshDespite all the recent development and success of deep neural networks, deployment of a deep model onto the resource-constrained devices still remains challenging. However, model pruning can resolve this issue for Convolutional Neural Networks (CNNs), since it is one of the most popular approaches to reducing computational complexities. Therefore, this article presents a pruning model for convolutional neural networks. The proposed method classifies and arranges similar filters into the same cluster where the similarity is calculated using a three-dimensional normalized cross-correlation. Moreover, these steps can be completed entirely based on the filter values while not requiring a set of test images as well as the acquisition of any filter activation. In the research, the performances of the proposed model pruning method have been evaluated, where it is observed that the proposed approach is computationally light and requires significantly less time and resources compared to ML and activation-based approaches. In the experiments, using the VGG16 model on the Cifar10 dataset, the proposed approach results in the pruned model(s) which are comparable in performance with models found using activation-based methods and expensive ML-based methods. Similar results are found when pruning a custom CNN on the MNIST and Fashion MNIST datasets as well.https://www.tandfonline.com/doi/10.1080/24751839.2024.2415008Convolutional neural networksdeep neural networkspruning
spellingShingle Niaz Ashraf Khan
A. M. Saadman Rafat
Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity
Journal of Information and Telecommunication
Convolutional neural networks
deep neural networks
pruning
title Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity
title_full Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity
title_fullStr Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity
title_full_unstemmed Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity
title_short Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity
title_sort pruning convolution neural networks using filter clustering based on normalized cross correlation similarity
topic Convolutional neural networks
deep neural networks
pruning
url https://www.tandfonline.com/doi/10.1080/24751839.2024.2415008
work_keys_str_mv AT niazashrafkhan pruningconvolutionneuralnetworksusingfilterclusteringbasedonnormalizedcrosscorrelationsimilarity
AT amsaadmanrafat pruningconvolutionneuralnetworksusingfilterclusteringbasedonnormalizedcrosscorrelationsimilarity