Filter pruning with uniqueness mechanism in the frequency domain for efficient neural networks

Zhang，Shuo1; Gao，Mingqi2,3; Ni，Qiang1; Han，Jungong4

2023-04-14

10.1016/j.neucom.2023.02.004

NEUROCOMPUTING   影响因子和分区

0925-2312

1872-8286

Filter pruning has drawn extensive attention due to its advantage in reducing computational costs and memory requirements of deep convolutional neural networks. However, most existing methods only prune filters based on their intrinsic properties or spatial feature maps, ignoring the correlation between filters. In this paper, we suggest the correlation is valuable and consider it from a novel view: the frequency domain. Specifically, we first transfer features to the frequency domain by Discrete Cosine Transform (DCT). Then, for each feature map, we compute a uniqueness score, which measures its probability of being replaced by others. This way allows to prune the filters corresponding to the low-uniqueness maps without significant performance degradation. Compared to the methods focusing on intrinsic properties, our proposed method introduces a more comprehensive criterion to prune filters, further improving the network compactness while preserving good performance. In addition, our method is more robust against noise than the spatial ones since the critical clues for pruning are more concentrated after DCT. Experimental results demonstrate the superiority of our method. To be specific, our method outperforms the baseline ResNet-56 by 0.38% on CIFAR-10 while reducing the floating-point operations (FLOPs) by 47.4%. In addition, a consistent improvement can be observed when pruning the baseline ResNet-110: 0.23% performance increase and up to 71% FLOPs drop. Finally, on ImageNet, our method reduces the FLOPs of the baseline ResNet-50 by 48.7% with only 0.32% accuracy loss.

Computer vision Deep learning Frequency-domain transformation Image classification Model compression

SCI ; EI

英语

其他

Computer Science

Computer Science, Artificial Intelligence

WOS:000947462600001

ELSEVIER

20230713591039

Computer vision ; Convolutional neural networks ; Deep neural networks ; Digital arithmetic ; Frequency domain analysis ; Image classification

Ergonomics and Human Factors Engineering:461.4 ; Computer Theory, Includes Formal Logic, Automata Theory, Switching Theory, Programming Theory:721.1 ; Data Processing and Image Processing:723.2 ; Computer Applications:723.5 ; Vision:741.2 ; Mathematical Transformations:921.3 ; Numerical Methods:921.6

COMPUTER SCIENCE

2-s2.0-85147913922

Scopus

1.School of Computing and Communications,Lancaster University,Lancaster,LA1 4WA,United Kingdom
2.WMG Data Science,The University of Warwick,Coventry,CV4 7AL,United Kingdom
3.Department of Computer Science and Engineering,Southern University of Science and Technology,Shenzhen,518055,China
4.Department of Computer Science,The University of Sheffield,Sheffield,Regent Court, 211 Portobello,S1 4DP,United Kingdom

Zhang，Shuo,Gao，Mingqi,Ni，Qiang,et al. Filter pruning with uniqueness mechanism in the frequency domain for efficient neural networks[J]. NEUROCOMPUTING,2023,530:116-124.

Zhang，Shuo,Gao，Mingqi,Ni，Qiang,&Han，Jungong.(2023).Filter pruning with uniqueness mechanism in the frequency domain for efficient neural networks.NEUROCOMPUTING,530,116-124.

Zhang，Shuo,et al."Filter pruning with uniqueness mechanism in the frequency domain for efficient neural networks".NEUROCOMPUTING 530(2023):116-124.