{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T18:37:34Z","timestamp":1782931054404,"version":"3.54.5"},"reference-count":51,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2023,3,2]],"date-time":"2023-03-02T00:00:00Z","timestamp":1677715200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100012166","name":"National Key Research and Development Program of China","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100012166","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Neurorobot."],"abstract":"<jats:p>Filter pruning is widely used for inference acceleration and compatibility with off-the-shelf hardware devices. Some filter pruning methods have proposed various criteria to approximate the importance of filters, and then sort the filters globally or locally to prune the redundant parameters. However, the current criterion-based methods have problems: (1) parameters with smaller criterion values for extracting edge features are easily ignored, and (2) there is a strong correlation between different criteria, resulting in similar pruning structures. In this article, we propose a novel simple but effective pruning method based on filter similarity, which is used to evaluate the similarity between filters instead of the importance of a single filter. The proposed method first calculates the similarity of the filters pairwise in one convolutional layer and then obtains the similarity distribution. Finally, the filters with high similarity to others are deleted from the distribution or set to zero. In addition, the proposed algorithm does not need to specify the pruning rate for each layer, and only needs to set the desired FLOPs or parameter reduction to obtain the final compression model. We also provide iterative pruning strategies for hard pruning and soft pruning to satisfy the tradeoff requirements of accuracy and memory in different scenarios. Extensive experiments on various representative benchmark datasets across different network architectures demonstrate the effectiveness of our proposed method. For example, on CIFAR10, the proposed algorithm achieves 61.1% FLOPs reduction by removing 58.3% of the parameters, with no loss in Top-1 accuracy on ResNet-56; and reduces 53.05% FLOPs on ResNet-50 with only 0.29% Top-1 accuracy degradation on ILSVRC-2012.<\/jats:p>","DOI":"10.3389\/fnbot.2023.1132679","type":"journal-article","created":{"date-parts":[[2023,3,2]],"date-time":"2023-03-02T04:55:22Z","timestamp":1677732922000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":6,"title":["Model pruning based on filter similarity for edge device deployment"],"prefix":"10.3389","volume":"17","author":[{"given":"Tingting","family":"Wu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chunhe","family":"Song","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Peng","family":"Zeng","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2023,3,2]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","author":"Ashok","year":"2017","DOI":"10.48550\/arXiv.1709.06030"},{"key":"B2","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v36i1.19888","article-title":"\u201cPrior gradient mask guided pruning-aware fine-tuning,\u201d","author":"Cai","year":"2022","journal-title":"Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 1"},{"key":"B3","first-page":"3123","article-title":"\u201cBinaryconnect: training deep neural networks with binary weights during propagations,\u201d","author":"Courbariaux","year":"2015","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B4","first-page":"1269","article-title":"\u201cExploiting linear structure within convolutional networks for efficient evaluation,\u201d","author":"Denton","year":"2014","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B5","doi-asserted-by":"publisher","first-page":"5840","DOI":"10.1109\/CVPR.2017.205","article-title":"\u201cMore is less: a more complicated network with less inference complexity,\u201d","author":"Dong","year":"2017","journal-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition"},{"key":"B6","doi-asserted-by":"publisher","first-page":"104298","DOI":"10.1016\/j.engappai.2021.104298","article-title":"Pushing artificial intelligence to the edge: emerging trends, issues and challenges","volume":"103","author":"Fortino","year":"2021","journal-title":"Eng. Appl. Artif. Intell"},{"key":"B7","doi-asserted-by":"publisher","first-page":"1789","DOI":"10.1007\/s11263-021-01453-z","article-title":"Knowledge distillation: a survey","volume":"129","author":"Gou","year":"2021","journal-title":"Int. J. Comput. Vis"},{"key":"B8","doi-asserted-by":"publisher","first-page":"6645","DOI":"10.1109\/ICASSP.2013.6638947","article-title":"\u201cSpeech recognition with deep recurrent neural networks,\u201d","author":"Graves","year":"2013","journal-title":"2013 IEEE International Conference on Acoustics, Speech and Signal Processing"},{"key":"B9","first-page":"164","article-title":"\u201cSecond order derivatives for network pruning: optimal brain surgeon,\u201d","author":"Hassibi","year":"1993","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B10","doi-asserted-by":"publisher","first-page":"770","DOI":"10.1109\/CVPR.2016.90","article-title":"\u201cDeep residual learning for image recognition,\u201d","author":"He","year":"2016","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition"},{"key":"B11","doi-asserted-by":"publisher","DOI":"10.24963\/ijcai.2018\/309","article-title":"Soft filter pruning for accelerating deep convolutional neural networks","author":"He","year":"","journal-title":"arXiv preprint arXiv:1808.06866"},{"key":"B12","doi-asserted-by":"publisher","first-page":"784","DOI":"10.1007\/978-3-030-01234-2_48","article-title":"\u201cAMC: Automl for model compression and acceleration on mobile devices,\u201d","author":"He","year":"","journal-title":"Proceedings of the European Conference on Computer Vision"},{"key":"B13","doi-asserted-by":"publisher","first-page":"4340","DOI":"10.1109\/CVPR.2019.00447","article-title":"\u201cFilter pruning via geometric median for deep convolutional neural networks acceleration,\u201d","author":"He","year":"2019","journal-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition"},{"key":"B14","doi-asserted-by":"publisher","first-page":"1389","DOI":"10.1109\/ICCV.2017.155","article-title":"\u201cChannel pruning for accelerating very deep neural networks,\u201d","author":"He","year":"2017","journal-title":"Proceedings of the IEEE International Conference on Computer Vision"},{"key":"B15","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1503.02531","article-title":"Distilling the knowledge in a neural network","author":"Hinton","year":"2015","journal-title":"arXiv preprint arXiv:1503.02531"},{"key":"B16","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1704.04861","article-title":"Mobilenets: efficient convolutional neural networks for mobile vision applications","author":"Howard","year":"2017","journal-title":"arXiv preprint arXiv:1704.04861"},{"key":"B17","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1607.03250","article-title":"Network trimming: a data-driven neuron pruning approach towards efficient deep architectures","author":"Hu","year":"2016","journal-title":"arXiv preprint arXiv:1607.03250"},{"key":"B18","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2004.11627","article-title":"Convolution-weight-distribution assumption: rethinking the criteria of channel pruning","author":"Huang","year":"2020","journal-title":"arXiv preprint arXiv:2004.11627"},{"key":"B19","doi-asserted-by":"publisher","first-page":"6869","DOI":"10.48550\/arXiv.1609.07061","article-title":"Quantized neural networks: training neural networks with low precision weights and activations","volume":"18","author":"Hubara","year":"2017","journal-title":"J. Mach. Learn. Res"},{"key":"B20","author":"Krizhevsky","year":"2009"},{"key":"B21","doi-asserted-by":"publisher","first-page":"1097","DOI":"10.1145\/3065386","article-title":"Imagenet classification with deep convolutional neural networks","volume":"25","author":"Krizhevsky","year":"2012","journal-title":"Adv. Neural Inform. Process. Syst"},{"key":"B22","first-page":"598","article-title":"\u201cOptimal brain damage,\u201d","author":"LeCun","year":"1990","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B23","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1810.02340","article-title":"SNIP: single-shot network pruning based on connection sensitivity","author":"Lee","year":"2018","journal-title":"arXiv preprint arXiv:1810.02340"},{"key":"B24","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1608.08710","article-title":"Pruning filters for efficient convnets","author":"Li","year":"2016","journal-title":"arXiv preprint arXiv:1608.08710"},{"key":"B25","first-page":"2181","article-title":"\u201cRuntime neural pruning,\u201d","author":"Lin","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B26","doi-asserted-by":"publisher","first-page":"1529","DOI":"10.1109\/CVPR42600.2020.00160","article-title":"\u201cHrank: filter pruning using high-rank feature map,\u201d","author":"Lin","year":"2020","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition"},{"key":"B27","first-page":"806","article-title":"\u201cSparse convolutional neural networks,\u201d","author":"Liu","year":"2015","journal-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition"},{"key":"B28","doi-asserted-by":"publisher","first-page":"2736","DOI":"10.1109\/ICCV.2017.298","article-title":"\u201cLearning efficient convolutional networks through network slimming,\u201d","author":"Liu","year":"2017","journal-title":"Proceedings of the IEEE International Conference on Computer Vision"},{"key":"B29","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1810.05270","article-title":"Rethinking the value of network pruning","author":"Liu","year":"2018","journal-title":"arXiv preprint arXiv:1810.05270"},{"key":"B30","doi-asserted-by":"publisher","first-page":"5058","DOI":"10.1109\/ICCV.2017.541","article-title":"\u201cThinet: a filter level pruning method for deep neural network compression,\u201d","author":"Luo","year":"2017","journal-title":"Proceedings of the IEEE International Conference on Computer Vision"},{"key":"B31","doi-asserted-by":"publisher","first-page":"2525","DOI":"10.1109\/TPAMI.2018.2858232","article-title":"Thinet: pruning CNN filters for a thinner net","volume":"41","author":"Luo","year":"2018","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell"},{"key":"B32","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1611.06440","article-title":"Pruning convolutional neural networks for resource efficient inference","author":"Molchanov","year":"2016","journal-title":"arXiv preprint arXiv:1611.06440"},{"key":"B33","first-page":"5113","article-title":"\u201cCollaborative channel pruning for deep networks,\u201d","author":"Peng","year":"2019","journal-title":"International Conference on Machine Learning"},{"key":"B34","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2016.2577031","article-title":"Faster R-CNN: towards real-time object detection with region proposal networks","author":"Ren","year":"2015","journal-title":"arXiv preprint arXiv:1506.01497"},{"key":"B35","doi-asserted-by":"publisher","first-page":"211","DOI":"10.1007\/s11263-015-0816-y","article-title":"Imagenet large scale visual recognition challenge","volume":"115","author":"Russakovsky","year":"2015","journal-title":"Int. J. Comput. Vis"},{"key":"B36","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1409.1556","article-title":"Very deep convolutional networks for large-scale image recognition","author":"Simonyan","year":"2014","journal-title":"arXiv preprint arXiv:1409.1556"},{"key":"B37","first-page":"1139","article-title":"\u201cOn the importance of initialization and momentum in deep learning,\u201d","author":"Sutskever","year":"2013","journal-title":"International Conference on Machine Learning"},{"key":"B38","doi-asserted-by":"publisher","first-page":"103775","DOI":"10.1016\/j.engappai.2020.103775","article-title":"Emotion recognition using speech and neural structured learning to facilitate edge intelligence","volume":"94","author":"Uddin","year":"2020","journal-title":"Eng. Appl. Artif. Intell"},{"key":"B39","doi-asserted-by":"publisher","first-page":"103785","DOI":"10.1016\/j.engappai.2020.103785","article-title":"Data flow and distributed deep neural network based low latency IoT-edge computation model for big data environment","volume":"94","author":"Veeramanikandan","year":"2020","journal-title":"Eng. Appl. Artif. Intell"},{"key":"B40","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1803.05729","article-title":"Exploring linear relationship in feature map subspace for convnets compression","author":"Wang","year":"2018","journal-title":"arXiv preprint arXiv:1803.05729"},{"key":"B41","doi-asserted-by":"publisher","first-page":"14913","DOI":"10.1109\/CVPR46437.2021.01467","article-title":"\u201cConvolutional neural network pruning with structural redundancy reduction,\u201d","author":"Wang","year":"2021","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition"},{"key":"B42","first-page":"2074","article-title":"\u201cLearning structured sparsity in deep neural networks,\u201d","author":"Wen","year":"2016","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B43","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1802.00124","article-title":"Rethinking the smaller-norm-less-informative assumption in channel pruning of convolution layers","author":"Ye","year":"2018","journal-title":"arXiv preprint arXiv:1802.00124"},{"key":"B44","doi-asserted-by":"publisher","first-page":"9194","DOI":"10.1109\/CVPR.2018.00958","article-title":"\u201cNISP: pruning networks using neuron importance score propagation,\u201d","author":"Yu","year":"2018","journal-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition"},{"key":"B45","doi-asserted-by":"publisher","DOI":"10.1109\/TMM.2022.3171088","article-title":"Incorporating linear regression problems into an adaptive framework with feasible optimizations","author":"Zhang","year":"2022","journal-title":"IEEE Trans. Multim"},{"key":"B46","doi-asserted-by":"publisher","first-page":"1722","DOI":"10.1109\/TCYB.2018.2811764","article-title":"LRR for subspace segmentation via tractable schatten-p norm minimization and factorization","volume":"49","author":"Zhang","year":"2018","journal-title":"IEEE Trans. Cybern"},{"key":"B47","doi-asserted-by":"publisher","first-page":"103774","DOI":"10.1016\/j.engappai.2020.103774","article-title":"Optimized task distribution based on task requirements and time delay in edge computing environments","volume":"94","author":"Zhang","year":"2020","journal-title":"Eng. Appl. Artif. Intell"},{"key":"B48","doi-asserted-by":"publisher","first-page":"6848","DOI":"10.1109\/CVPR.2018.00716","article-title":"\u201cShuffleNet: an extremely efficient convolutional neural network for mobile devices,\u201d","author":"Zhang","year":"2018","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition"},{"key":"B49","doi-asserted-by":"publisher","first-page":"1626","DOI":"10.1109\/TCYB.2019.2928174","article-title":"A knee-guided evolutionary algorithm for compressing deep neural networks","volume":"51","author":"Zhou","year":"2019","journal-title":"IEEE Trans. Cybern"},{"key":"B50","first-page":"875","article-title":"\u201cDiscrimination-aware channel pruning for deep neural networks,\u201d","author":"Zhuang","year":"2018","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B51","article-title":"SCSP: spectral clustering filter pruning with soft self-adaption manners","author":"Zhuo","year":"2018","journal-title":"arXiv preprint arXiv:1806.05320"}],"container-title":["Frontiers in Neurorobotics"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fnbot.2023.1132679\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,3,2]],"date-time":"2023-03-02T04:55:40Z","timestamp":1677732940000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fnbot.2023.1132679\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,3,2]]},"references-count":51,"alternative-id":["10.3389\/fnbot.2023.1132679"],"URL":"https:\/\/doi.org\/10.3389\/fnbot.2023.1132679","relation":{},"ISSN":["1662-5218"],"issn-type":[{"value":"1662-5218","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,3,2]]},"article-number":"1132679"}}