Filter Pruning via Learned Representation Median in the Frequency Domain.

Zhang, X; Xie, W; Li, Y; Lei, J; Du, Q

Filter Pruning via Learned Representation Median in the Frequency Domain.

Zhang, X Xie, W Li, Y Lei, J

Du, Q

Permalink

Publisher:: IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
Publication Type:: Journal Article
Citation:: IEEE Trans Cybern, 2023, 53, (5), pp. 3165-3175
Issue Date:: 2023-05

Closed Access

	Filename	Description	Size
	Filter_Pruning_via_Learned_Representation_Median_in_the_Frequency_Domain.pdf	Published version	6.72 MB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Zhang, X
dc.contributor.author	Xie, W
dc.contributor.author	Li, Y
dc.contributor.author	Lei, J https://orcid.org/0000-0003-0851-6565
dc.contributor.author	Du, Q
dc.date.accessioned	2024-04-29T09:23:56Z
dc.date.available	2024-04-29T09:23:56Z
dc.date.issued	2023-05
dc.identifier.citation	IEEE Trans Cybern, 2023, 53, (5), pp. 3165-3175
dc.identifier.issn	2168-2267
dc.identifier.issn	2168-2275
dc.identifier.uri	http://hdl.handle.net/10453/178481
dc.description.abstract	In this article, we propose a novel filter pruning method for deep learning networks by calculating the learned representation median (RM) in frequency domain (LRMF). In contrast to the existing filter pruning methods that remove relatively unimportant filters in the spatial domain, our newly proposed approach emphasizes the removal of absolutely unimportant filters in the frequency domain. Through extensive experiments, we observed that the criterion for "relative unimportance" cannot be generalized well and that the discrete cosine transform (DCT) domain can eliminate redundancy and emphasize low-frequency representation, which is consistent with the human visual system. Based on these important observations, our LRMF calculates the learned RM in the frequency domain and removes its corresponding filter, since it is absolutely unimportant at each layer. Thanks to this, the time-consuming fine-tuning process is not required in LRMF. The results show that LRMF outperforms state-of-the-art pruning methods. For example, with ResNet110 on CIFAR-10, it achieves a 52.3% FLOPs reduction with an improvement of 0.04% in Top-1 accuracy. With VGG16 on CIFAR-100, it reduces FLOPs by 35.9% while increasing accuracy by 0.5%. On ImageNet, ResNet18 and ResNet50 are accelerated by 53.3% and 52.7% with only 1.76% and 0.8% accuracy loss, respectively. The code is based on PyTorch and is available at https://github.com/zhangxin-xd/LRMF.
dc.format	Print-Electronic
dc.language	eng
dc.publisher	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
dc.relation.ispartof	IEEE Trans Cybern
dc.relation.isbasedon	10.1109/TCYB.2021.3124284
dc.rights	info:eu-repo/semantics/closedAccess
dc.title	Filter Pruning via Learned Representation Median in the Frequency Domain.
dc.type	Journal Article
utslib.citation.volume	53
utslib.location.activity	United States
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology/School of Electrical and Data Engineering
utslib.copyright.status	closed_access	*
dc.date.updated	2024-04-29T09:23:52Z
pubs.issue	5
pubs.publication-status	Published
pubs.volume	53
utslib.citation.issue	5

Abstract:

In this article, we propose a novel filter pruning method for deep learning networks by calculating the learned representation median (RM) in frequency domain (LRMF). In contrast to the existing filter pruning methods that remove relatively unimportant filters in the spatial domain, our newly proposed approach emphasizes the removal of absolutely unimportant filters in the frequency domain. Through extensive experiments, we observed that the criterion for "relative unimportance" cannot be generalized well and that the discrete cosine transform (DCT) domain can eliminate redundancy and emphasize low-frequency representation, which is consistent with the human visual system. Based on these important observations, our LRMF calculates the learned RM in the frequency domain and removes its corresponding filter, since it is absolutely unimportant at each layer. Thanks to this, the time-consuming fine-tuning process is not required in LRMF. The results show that LRMF outperforms state-of-the-art pruning methods. For example, with ResNet110 on CIFAR-10, it achieves a 52.3% FLOPs reduction with an improvement of 0.04% in Top-1 accuracy. With VGG16 on CIFAR-100, it reduces FLOPs by 35.9% while increasing accuracy by 0.5%. On ImageNet, ResNet18 and ResNet50 are accelerated by 53.3% and 52.7% with only 1.76% and 0.8% accuracy loss, respectively. The code is based on PyTorch and is available at https://github.com/zhangxin-xd/LRMF.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/178481