Abstract
Filter pruning is one of the most effective approaches to reduce the storage and computational cost of convolutional neural networks. How to measure the importance of each filter is the key problem for filter pruning. In this work, we propose a novel method that can evaluate the importance of each filter and gradually prunes those filters with small scores. Specifically, the importance is obtained via probing the effect of each filter on the task-related loss function by randomly pruning the original network. The smaller the effect on the task-related loss function, the lower the importance of the filter. It’s worth noting that our method is scale consistent across all layers without requiring layer-wise sensitivity analysis, which can be used to prune various networks, including ResNet and DenseNet. Extensive experiments demonstrate the outstanding performance of our method. For example, on ILSVRC-2012, our method can prune 42.74% floating point operations and 39.61% parameters of ResNet-50 with only 0.73% Top-1 accuracy loss and 0.37% Top-5 accuracy loss.
Similar content being viewed by others
References
Chen LC, Papandreou G, Kokkinos I, Murphy K, Yuille AL (2015) Semantic image segmentation with deep convolutional nets and fully connected crfs. In: Int. conf. on learning representations (ICLR)
Chen P, Liu S, Zhao H, Jia J (2021) Distilling knowledge via knowledge review. In: Proc. IEEE conf. computer vision and pattern recognition, pp 5008–5017
Chollet F (2017) Xception: deep learning with depthwise separable convolutions. In: Proc. IEEE conf. computer vision and pattern recognition, pp 1800–1807
Chu X, Zhang B, Xu R, Li J (2019) Fairnas: rethinking evaluation fairness of weight sharing neural architecture search. arXiv preprint arXiv:1907.01845
Courbariaux M, Bengio Y, David JP (2015) Binaryconnect: training deep neural networks with binary weights during propagations. Adv Neural Inf Process Syst 28:3123–3131
Deng J, Dong W, Socher R, Li LJ, Li K, Fei-Fei L (2009) Imagenet: a large-scale hierarchical image database. In: Proc. IEEE conf. computer vision and pattern recognition, pp 248–255
Ding X, Ding G, Guo Y, Han J, Yan C (2019) Approximated oracle filter pruning for destructive cnn width optimization. In: Int. conf. on machine learning (ICML), pp 1607–1616
Dong X, Yang Y (2019) Network pruning via transformable architecture search. Adv Neural Inf Process Syst 32:760–771
Girshick R, Donahue J, Darrell T, Malik J (2014) Rich feature hierarchies for accurate object detection and semantic segmentation. In: Proc. IEEE conf. computer vision and pattern recognition, pp 580–587
Guo Y, Yao A, Chen Y (2016) Dynamic network surgery for efficient dnns. Adv Neural Inf Process Syst 29:1387–1395
Guo Z, Zhang X, Mu H, Heng W, Liu Z, Wei Y, Sun J (2020) Single path one-shot neural architecture search with uniform sampling. In: Proc. European conf. computer vision (ECCV), pp 544–560
Han S, Pool J, Tran J, Dally WJ (2015) Learning both weights and connections for efficient neural networks. Adv Neural Inf Process Syst 28:1135–1143
Han S, Mao H, Dally WJ (2016) Deep compression: compressing deep neural networks with pruning, trained quantization and Huffman coding. In: Int. conf. on learning representations (ICLR)
He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. In: Proc. IEEE conf. computer vision and pattern recognition, pp 770–778
He Y, Zhang X, Sun J (2017) Channel pruning for accelerating very deep neural networks. In: The IEEE int. conf. computer vision (ICCV), pp 1398–1406
He Y, Kang G, Dong X, Fu Y, Yang Y (2018) Soft filter pruning for accelerating deep convolutional neural networks. In: Int. joint conf. artificial intelligence (IJCAI), pp 2234–2240
He Y, Lin J, Liu Z, Wang H, Li LJ, Han S (2018) Amc: automl for model compression and acceleration on mobile devices. In: Proc. European conf. computer vision (ECCV), pp 784–800
He Y, Liu P, Wang Z, Hu Z, Yang Y (2019) Filter pruning via geometric median for deep convolutional neural networks acceleration. In: Proc. IEEE conf. computer vision and pattern recognition, pp 4340–4349
He Y, Ding Y, Liu P, Zhu L, Zhang H, Yang Y (2020) Learning filter pruning criteria for deep convolutional neural networks acceleration. In: Proc. IEEE conf. computer vision and pattern recognition, pp –2018
Hinton GE, Vinyals O, Dean J (2015) Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531
Huang G, Liu Z, van der Maaten L, Weinberger KQ (2017) Densely connected convolutional networks. In: Proc. IEEE conf. computer vision and pattern recognition, pp 2261–2269
Hubara I, Courbariaux M, Soudry D, El-Yaniv R, Bengio Y (2017) Quantized neural networks: training neural networks with low precision weights and activations. J Mach Learn Res 18(1):6869–6898
Hu H, Peng R, Tai YW, Tang CK (2016) Network trimming: a data-driven neuron pruning approach towards efficient deep architectures. arXiv preprint arXiv:1607.03250
Jaderberg M, Vedaldi A, Zisserman A (2014) Speeding up convolutional neural networks with low rank expansions. In: British machine vision conference (BMVC)
Krizhevsky A (2009) Learning multiple layers of features from tiny images. In: Technical report
Krizhevsky A, Sutskever I, Hinton GE (2017) Imagenet classification with deep convolutional neural networks. Commun ACM 60(6):84–90
Li H, Kadav A, Durdanovic I, Samet H, Graf HP (2017) Pruning filters for efficient convnets. In: Int. conf. on learning representations (ICLR)
Li B, Wu B, Su J, Wang G, Lin L (2020) Eagleeye: fast sub-net evaluation for efficient neural network pruning. In: Proc. European conf. computer vision (ECCV), pp 639–654
Li Y, Gong R, Tan X, Yang Y, Hu P, Zhang Q, Yu F, Wang W, Gu S (2021) Brecq: pushing the limit of post-training quantization by block reconstruction. In: Int. conf. on learning representations (ICLR)
Lin S, Ji R, Yan C, Zhang B, Cao L, Ye Q, Huang F, Doermann D (2019) Towards optimal structured cnn pruning via generative adversarial learning. In: Proc. IEEE conf. computer vision and pattern recognition, pp 2790–2799
Lin M, Ji R, Wang Y, Zhang Y, Zhang B, Tian Y, Shao L (2020) Hrank: Filter pruning using high-rank feature map. In: Proc. IEEE conf. computer vision and pattern recognition, pp 1529–1538
Liu Z, Li J, Shen Z, Huang G, Yan S, Zhang C (2017) Learning efficient convolutional networks through network slimming. In: The IEEE int. conf. computer vision (ICCV), pp 2755–2763
Liu Z, Mu H, Zhang X, Guo Z, Yang X, Cheng KT, Sun J (2019) Metapruning: meta learning for automatic neural network channel pruning. In: The IEEE int. conf. computer vision (ICCV), pp 3296–3305
Liu H, Simonyan K, Yang Y (2019) Darts: differentiable architecture search. In: Int. conf. on learning representations (ICLR)
Luo JH, Wu J, Lin W (2017) Thinet: a filter level pruning method for deep neural network compression. In: The IEEE int. conf. computer vision (ICCV), pp 5068–5076
Molchanov P, Tyree S, Karras T, Aila T, Kautz J (2017) Pruning convolutional neural networks for resource efficient inference. In: Int. conf. on learning representations (ICLR)
Ning X, Zhao T, Li W, Lei P, Wang Y, Yang H (2020) Dsa: more efficient budgeted pruning via differentiable sparsity allocation. In: Proc. European conf. computer vision (ECCV), pp 592–607
Petsiuk V, Das A, Saenko K (2018) Rise: randomized input sampling for explanation of black-box models. In: British machine vision conference (BMVC), p 151
Ren S, He K, Girshick R, Sun J (2017) Faster r-cnn: towards real-time object detection with region proposal networks. IEEE Trans Pattern Anal Mach Intell 39(6):1137–1149
Shao M, Dai J, Kuang J, Meng D (2021) A dynamic CNN pruning method based on matrix similarity. Signal Image Video Process 15(2):381–389. https://doi.org/10.1007/s11760-020-01760-x
Shao M, Dai J, Wang R, Zuo W (2021) CSHE: network pruning by using cluster similarity and matrix eigenvalues. Int J Mach Learn Cybernet. https://doi.org/10.1007/s13042-021-01411-8
Simonyan K, Zisserman A (2015) Very deep convolutional networks for large-scale image recognition. In: Int. conf. on learning representations (ICLR)
Tang Y, Wang Y, Xu Y, Deng Y, Xu C, Tao D, Xu C (2021) Manifold regularized dynamic network pruning. In: Proc. IEEE conf. computer vision and pattern recognition, pp 5018–5028
Wen W, Wu C, Wang Y, Chen Y, Li H (2016) Learning structured sparsity in deep neural networks. Adv Neural Inf Process Syst 29:2074–2082
Yamamoto K (2021) Learnable companding quantization for accurate low-bit neural networks. In: Proc. IEEE conf. computer vision and pattern recognition, pp 5029–5038
Ye J, Lu X, Lin ZL, Wang JZ (2018) Rethinking the smaller-norm-less-informative assumption in channel pruning of convolution layers. In: Int. conf. on learning representations (ICLR)
You Z, Yan K, Ye J, Ma M, Wang P (2019) Gate decorator: global filter pruning method for accelerating deep convolutional neural networks. Adv Neural Inf Process Syst 32:2133–2144
Yu J, Huang T (2019) Universally slimmable networks and improved training techniques. In: The ieee int. conf. computer vision (ICCV), pp 1803–1811
Zhao C, Ni B, Zhang J, Zhao Q, Zhang W, Tian Q (2019) Variational convolutional neural network pruning. In: Proc. IEEE conf. computer vision and pattern recognition, pp 2780–2789
Zhuang Z, Tan M, Zhuang B, Liu J, Guo Y, Wu Q, Huang J, Zhu J (2018) Discrimination-aware channel pruning for deep neural networks. Adv Neural Inf Process Syst 31:883–894
Acknowledgements
The authors are very indebted to the anonymous referees for their critical comments and suggestions for the improvement of this paper. This work was supported by National Key Research and development Program of China (2021YFA1000102), and in part by the grants from the National Natural Science Foundation of China (nos. 61673396, 61976245, 61772344).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Kuang, J., Shao, M., Wang, R. et al. Network pruning via probing the importance of filters. Int. J. Mach. Learn. & Cyber. 13, 2403–2414 (2022). https://doi.org/10.1007/s13042-022-01530-w
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13042-022-01530-w