{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,19]],"date-time":"2026-03-19T06:35:31Z","timestamp":1773902131271,"version":"3.50.1"},"reference-count":65,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,4,17]],"date-time":"2025-04-17T00:00:00Z","timestamp":1744848000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Comput. Sci."],"abstract":"<jats:sec><jats:title>Introduction<\/jats:title><jats:p>In recent years, Deep Learning (DL) and Artificial Neural Networks (ANNs) have transformed industrial applications by providing automation in complex tasks such as anomaly detection and predictive maintenance. However, traditional DL models often need significant computational resources, making them unsuitable for resource-constrained edge devices. This paper explores the potential of sparse ANNs to address these challenges, focusing on their application in industrial settings.<\/jats:p><\/jats:sec><jats:sec><jats:title>Methods<\/jats:title><jats:p>We perform an experimental comparison of pruning techniques, including the Pre-Training, In-Training, Post-Training, and SET method, applied to the VGG16 and ResNet18 architectures, and conduct a systematic analysis of pruning methodologies alongside the effects of varying sparsity levels, to analyze their efficiency in anomaly detection and object classification tasks. Key metrics such as training accuracy, inference time, and energy consumption are analyzed to assess the feasibility of deploying sparse models on edge devices.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results and discussion<\/jats:title><jats:p>Our results demonstrate that sparse ANNs, particularly when pruned using the SET method, achieve energy savings without compromising accuracy, making them suitable for industrial applications. This work highlights the potential of sparse neural networks to boost sustainability and efficiency in industrial environments, paving the way for their large adoption in edge computing scenarios.<\/jats:p><\/jats:sec>","DOI":"10.3389\/fcomp.2025.1563942","type":"journal-article","created":{"date-parts":[[2025,4,17]],"date-time":"2025-04-17T05:24:06Z","timestamp":1744867446000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":2,"title":["A comparative study of neural network pruning strategies for industrial applications"],"prefix":"10.3389","volume":"7","author":[{"given":"Amirhossein","family":"Douzandeh Zenoozi","sequence":"first","affiliation":[]},{"given":"Laura","family":"Erhan","sequence":"additional","affiliation":[]},{"given":"Antonio","family":"Liotta","sequence":"additional","affiliation":[]},{"given":"Lucia","family":"Cavallaro","sequence":"additional","affiliation":[]}],"member":"1965","published-online":{"date-parts":[[2025,4,17]]},"reference":[{"key":"B1","doi-asserted-by":"crossref","first-page":"1706","DOI":"10.1109\/ICIP46576.2022.9897283","article-title":"\u201cAnomalib: a deep learning library for anomaly detection,\u201d","volume-title":"2022 IEEE International Conference on Image Processing (ICIP)","author":"Akcay","year":"2022"},{"key":"B2","doi-asserted-by":"publisher","first-page":"71","DOI":"10.3390\/informatics11040071","article-title":"Edge computing and cloud computing for internet of things: a review","volume":"11","author":"Andriulo","year":"2024","journal-title":"Informatics (MDPI)"},{"key":"B3","doi-asserted-by":"publisher","first-page":"10263","DOI":"10.1007\/s12652-021-03688-7","article-title":"Industrial object and defect recognition utilizing multilevel feature extraction from industrial scenes with deep learning approach","volume":"14","author":"Apostolopoulos","year":"2022","journal-title":"J. Ambient Intell. Humaniz. Comput"},{"key":"B4","article-title":"Unveiling the power of sparse neural networks for feature selection","author":"Atashgahi","year":"2024","journal-title":"arXiv:2408.04583."},{"key":"B5","doi-asserted-by":"crossref","first-page":"33","DOI":"10.1145\/3295500.3356202","article-title":"\u201cScalable reinforcement-learning-based neural architecture search for cancer deep learning research,\u201d","volume-title":"Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC '19","author":"Balaprakash","year":"2019"},{"key":"B6","author":"Blalock","year":"2020"},{"key":"B7","doi-asserted-by":"crossref","first-page":"674","DOI":"10.1109\/MWSCAS60917.2024.10658964","article-title":"\u201cExtremely low-power edge connected devices,\u201d","volume-title":"2024 IEEE 67th International Midwest Symposium on Circuits and Systems (MWSCAS)","author":"Brennan","year":"2024"},{"key":"B8","article-title":"An analysis of deep neural network models for practical applications","author":"Canziani","year":"2016","journal-title":"arXiv:1605.07678"},{"key":"B9","article-title":"APRIL-GAN: a zero-\/few-shot anomaly classification and segmentation method for CVPR 2023 VAND workshop challenge tracks 1&2: 1st place on zero-shot AD and 4th place on few-shot AD","author":"Chen","year":"2023","journal-title":"arXiv:2305.17382"},{"key":"B10","doi-asserted-by":"publisher","first-page":"10558","DOI":"10.1109\/TPAMI.2024.3447085","article-title":"A survey on deep neural network pruning: taxonomy, comparison, analysis, and recommendations","volume":"46","author":"Cheng","year":"2024","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell"},{"key":"B11","doi-asserted-by":"crossref","first-page":"248","DOI":"10.1109\/CVPR.2009.5206848","article-title":"\u201cImagenet: a large-scale hierarchical image database,\u201d","volume-title":"2009 IEEE Conference on Computer Vision and Pattern Recognition","author":"Deng","year":"2009"},{"key":"B12","doi-asserted-by":"publisher","first-page":"7669","DOI":"10.1038\/s41598-025-92156-9","article-title":"Rethinking model prototyping through the MedMNIST+ dataset collection","volume":"15","author":"Doerrich","year":"2024","journal-title":"Sci. Rep"},{"key":"B13","doi-asserted-by":"publisher","first-page":"2200","DOI":"10.1109\/ICCVW.2017.257","article-title":"\u201cIntroducing mvtec itodd \u2013 a dataset for 3d object recognition in industry,\u201d","author":"Drost","year":"2017","journal-title":"2017 IEEE International Conference on Computer Vision Workshops (ICCVW)"},{"key":"B14","doi-asserted-by":"publisher","first-page":"1036","DOI":"10.1177\/1077546320936506","article-title":"A novel ResNet-based model structure and its applications in machine health monitoring","volume":"27","author":"Duan","year":"2021","journal-title":"J. Vib. Control"},{"key":"B15","doi-asserted-by":"publisher","first-page":"290","DOI":"10.5486\/PMD.1959.6.3-4.12","article-title":"On random graphs","volume":"6","author":"Erd\u0151s","year":"2022","journal-title":"I. Publ. Math. Debrecen"},{"key":"B16","first-page":"101","article-title":"\u201cComparing training of sparse to classic neural networks for binary classification in medical data,\u201d","volume-title":"Lecture Notes in Computer Science","author":"Erhan","year":"2025"},{"key":"B17","article-title":"Why random pruning is all we need to start sparse","author":"Gadhikar","year":"2022","journal-title":"arXiv:2210.02412"},{"key":"B18","doi-asserted-by":"publisher","first-page":"20092","DOI":"10.1109\/ACCESS.2022.3147838","article-title":"Detection of wheat unsound kernels based on improved ResNet","volume":"10","author":"Gao","year":"2022","journal-title":"IEEE Access"},{"key":"B19","article-title":"Generative adversarial networks","author":"Goodfellow","year":"2014","journal-title":"arXiv:1406.2661"},{"key":"B20","doi-asserted-by":"crossref","first-page":"770","DOI":"10.1109\/CVPR.2016.90","article-title":"\u201cDeep residual learning for image recognition,\u201d","volume-title":"2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)","author":"He","year":"2016"},{"key":"B21","doi-asserted-by":"publisher","first-page":"1027","DOI":"10.1109\/TII.2018.2875149","article-title":"Automatic fruit classification using deep learning for industrial applications","volume":"15","author":"Hossain","year":"2019","journal-title":"IEEE Trans. Industr. Inform"},{"key":"B22","doi-asserted-by":"publisher","first-page":"7883","DOI":"10.1609\/aaai.v35i9.16962","article-title":"Large batch optimization for deep learning using new complete layer-wise adaptive rate scaling","volume":"35","author":"Huo","year":"2021","journal-title":"Proc. Conf. AAAI Artif. Intell"},{"key":"B23","doi-asserted-by":"publisher","first-page":"110671","DOI":"10.1016\/j.patcog.2024.110671","article-title":"Resprune: an energy-efficient restorative filter pruning method using stochastic optimization for accelerating CNN","volume":"155","author":"Jayasimhan","year":"2024","journal-title":"Pattern Recogn"},{"key":"B24","doi-asserted-by":"crossref","first-page":"19606","DOI":"10.1109\/CVPR52729.2023.01878","article-title":"\u201cWinclip: Zero-\/few-shot anomaly classification and segmentation,\u201d","volume-title":"2023 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)","author":"Jeong","year":"2023"},{"key":"B25","doi-asserted-by":"publisher","first-page":"1","DOI":"10.2139\/ssrn.5012466","article-title":"Edge computing and analytics for iot devices: enhancing real-time decision making in smart environments","volume":"6","author":"Kanagarla","year":"2024","journal-title":"Int. J. Multidisc. Res"},{"key":"B26","doi-asserted-by":"publisher","first-page":"2704","DOI":"10.3390\/electronics13142704","article-title":"Handling power depletion in energy harvesting IoT devices","volume":"13","author":"Kang","year":"2024","journal-title":"Electronics"},{"key":"B27","doi-asserted-by":"crossref","first-page":"239","DOI":"10.1109\/ICCWAMTIP.2017.8301487","article-title":"\u201cAnalysis of resnet model for malicious code detection,\u201d","volume-title":"2017 14th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","author":"Khan","year":"2017"},{"key":"B28","doi-asserted-by":"publisher","first-page":"023","DOI":"10.30574\/ijsra.2024.13.2.2082","article-title":"Edge computing: revolutionizing data processing for IoT applications","volume":"13","author":"Kolapo","year":"2024","journal-title":"Int. J. Sci. Res. Arch"},{"key":"B29","article-title":"One weird trick for parallelizing convolutional neural networks","author":"Krizhevsky","year":"2014","journal-title":"arXiv:1404.5997"},{"key":"B30","article-title":"Dynamic sparse training with structured sparsity","author":"Lasby","year":"2023","journal-title":"arXiv:2305.02299"},{"key":"B31","article-title":"Minimizing energy consumption of deep learning models by energy-aware training","author":"Lazzaro","year":"2023","journal-title":"arXiv:2307.00368."},{"key":"B32","doi-asserted-by":"publisher","first-page":"357","DOI":"10.1016\/j.jmsy.2020.10.013","article-title":"Fault detection based on one-class deep learning for manufacturing applications limited to an imbalanced database","volume":"57","author":"Lee","year":"2020","journal-title":"J. Manuf. Syst"},{"key":"B33","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/ISCAS58744.2024.10558072","article-title":"\u201cDynamic gradient sparse update for edge training,\u201d","author":"Li","year":"2024","journal-title":"2024 IEEE International Symposium on Circuits and Systems (ISCAS)"},{"key":"B34","doi-asserted-by":"publisher","first-page":"370","DOI":"10.1016\/j.neucom.2021.07.045","article-title":"Pruning and quantization for deep neural network acceleration: a survey","volume":"461","author":"Liang","year":"2021","journal-title":"Neurocomputing"},{"key":"B35","article-title":"Microsoft COCO: common objects in context","author":"Lin","year":"2014","journal-title":"arXiv:1405.0312."},{"key":"B36","article-title":"Effective model sparsification by scheduled grow-and-prune methods","author":"Ma","year":"2021","journal-title":"arXiv:2106.09857"},{"key":"B37","doi-asserted-by":"publisher","first-page":"203","DOI":"10.1007\/s00138-018-01001-9","article-title":"Studying the plasticity in deep convolutional neural networks using random pruning","volume":"30","author":"Mittal","year":"2019","journal-title":"Mach. Vis. Appl"},{"key":"B38","doi-asserted-by":"publisher","first-page":"2383","DOI":"10.1038\/s41467-018-04316-3","article-title":"Scalable training of artificial neural networks with adaptive sparse connectivity inspired by network science","volume":"9","author":"Mocanu","year":"2018","journal-title":"Nat. Commun"},{"key":"B39","article-title":"SparseProp: Efficient sparse backpropagation for faster training of neural networks","author":"Nikdan","year":"2023","journal-title":"arXiv:2302.04852"},{"key":"B40","doi-asserted-by":"crossref","first-page":"57","DOI":"10.1002\/9781394219230.ch4","article-title":"\u201cModel optimization techniques for edge devices,\u201d","volume-title":"Model Optimization Methods for Efficient and Edge AI, Ch. 4","author":"Nimmagadda","year":"2025"},{"key":"B41","doi-asserted-by":"publisher","first-page":"110823","DOI":"10.1016\/j.comnet.2024.110823","article-title":"Edge-device collaborative computing for multi-view classification","volume":"254","author":"Palena","year":"2024","journal-title":"Comput. Netw"},{"key":"B42","article-title":"Sparser training for on-device recommendation systems","author":"Qu","year":"2024","journal-title":"arXiv:2411.12205"},{"key":"B43","doi-asserted-by":"crossref","first-page":"39","DOI":"10.70593\/978-81-981271-0-5_2","article-title":"\u201cScalable and adaptive deep learning algorithms for large-scale machine learning systems,\u201d","volume-title":"Future Research Opportunities for Artificial Intelligence in Industry 4.0 and 5.0","author":"Rane","year":"2024"},{"key":"B44","doi-asserted-by":"crossref","first-page":"779","DOI":"10.1109\/CVPR.2016.91","article-title":"\u201cYou only look once: unified, real-time object detection,\u201d","volume-title":"2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)","author":"Redmon","year":"2016"},{"key":"B45","doi-asserted-by":"publisher","first-page":"740","DOI":"10.1109\/72.248452","article-title":"Pruning algorithms-a survey","volume":"4","author":"Reed","year":"1993","journal-title":"IEEE Trans. Neural Netw"},{"key":"B46","doi-asserted-by":"crossref","first-page":"2199","DOI":"10.3850\/978-981-18-8071-1_P400-cd","article-title":"\u201cSynthiCAD: generation of industrial image data sets for resilience evaluation of safety-critical classifiers,\u201d","volume-title":"Proceeding of the 33rd European Safety and Reliability Conference","author":"Schuerrle","year":"2023"},{"key":"B47","doi-asserted-by":"publisher","first-page":"127280","DOI":"10.1016\/j.neucom.2024.127280","article-title":"Sparse optimization guided pruning for neural networks","volume":"574","author":"Shi","year":"2024","journal-title":"Neurocomputing"},{"key":"B48","doi-asserted-by":"publisher","first-page":"42","DOI":"10.1109\/JPROC.2022.3226481","article-title":"Efficient acceleration of deep learning inference on resource-constrained edge devices: a review","volume":"111","author":"Shuvo","year":"2023","journal-title":"Proceed. IEEE"},{"key":"B49","article-title":"Very deep convolutional networks for large-scale image recognition","author":"Simonyan","year":"2014","journal-title":"arXiv:1409.1556"},{"key":"B50","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/MOCAST61810.2024.10615490","article-title":"\u201cEnergy consumption aspects on embedded system for IoT applications,\u201d","volume-title":"2024 13th International Conference on Modern Circuits and Systems Technologies (MOCAST)","author":"Sofianidis","year":"2024"},{"key":"B51","doi-asserted-by":"publisher","first-page":"6904","DOI":"10.11591\/ijece.v13i6.pp6904-6912","article-title":"Compact optimized deep learning model for edge: a review","volume":"13","author":"Soumyalatha","year":"2023","journal-title":"Int. J. Electr. Comput. Eng"},{"key":"B52","doi-asserted-by":"publisher","first-page":"e1564","DOI":"10.7717\/peerj-cs.1564","article-title":"Random pruning: channel sparsity by expectation scaling factor","volume":"9","author":"Sun","year":"2023","journal-title":"PeerJ Comput. Sci"},{"key":"B53","article-title":"ULEEN: a novel architecture for ultra low-energy edge neural networks","author":"Susskind","year":"2023","journal-title":"arXiv:2304.10618."},{"key":"B54","doi-asserted-by":"crossref","first-page":"787","DOI":"10.23919\/FRUCT64283.2024.10749961","article-title":"\u201cThe convergence of edge computing and IoT-A paradigm shift in data processing,\u201d","volume-title":"2024 36th Conference of Open Innovations Association (FRUCT)","author":"Taha","year":"2024"},{"key":"B55","doi-asserted-by":"publisher","DOI":"10.36227\/techrxiv.172720330.08872446\/v1","article-title":"Improving neural network efficiency through advanced pruning techniques","author":"Tyche","year":"2024","journal-title":"Techrxiv"},{"key":"B56","doi-asserted-by":"publisher","first-page":"152","DOI":"10.54254\/2755-2721\/19\/20231025","article-title":"Research on pruning optimization techniques for neural networks","volume":"19","author":"Wang","year":"2023","journal-title":"Appl. Comput. Eng"},{"key":"B57","doi-asserted-by":"crossref","first-page":"359","DOI":"10.1007\/978-3-031-64832-8_19","article-title":"\u201cEnergy-efficient AI on the edge,\u201d","volume-title":"Unlocking Artificial Intelligence","author":"Witt","year":"2024"},{"key":"B58","doi-asserted-by":"publisher","first-page":"2043","DOI":"10.1109\/TNNLS.2018.2876179","article-title":"L1 -norm batch normalization for efficient training of deep neural networks","volume":"30","author":"Wu","year":"2019","journal-title":"IEEE Trans. Neural Netw. Learn. Syst"},{"key":"B59","doi-asserted-by":"publisher","first-page":"107707","DOI":"10.1016\/j.asoc.2021.107707","article-title":"A novel weight pruning strategy for light weight neural networks with application to the diagnosis of skin disease","volume":"111","author":"Xiang","year":"2021","journal-title":"Appl. Soft Comput"},{"key":"B60","doi-asserted-by":"publisher","first-page":"107660","DOI":"10.1016\/j.cmpb.2023.107660","article-title":"ResNet and its application to medical image processing: research progress and challenges","volume":"240","author":"Xu","year":"2023","journal-title":"Comput. Methods Progr. Biomed"},{"key":"B61","doi-asserted-by":"crossref","first-page":"191","DOI":"10.1109\/ISBI48211.2021.9434062","article-title":"\u201cMedmnist classification decathlon: a lightweight automl benchmark for medical image analysis,\u201d","volume-title":"2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI)","author":"Yang","year":"2021"},{"key":"B62","doi-asserted-by":"publisher","first-page":"41","DOI":"10.1038\/s41597-022-01721-8","article-title":"Medmnist v2-a large-scale lightweight benchmark for 2d and 3d biomedical image classification","volume":"10","author":"Yang","year":"2023","journal-title":"Sci. Data"},{"key":"B63","first-page":"119","article-title":"\u201cZeus: understanding and optimizing GPU energy consumption of DNN training,\u201d","volume-title":"20th USENIX Symposium on Networked Systems Design and Implementation (NSDI 23)","author":"You","year":"2023"},{"key":"B64","doi-asserted-by":"publisher","first-page":"232","DOI":"10.54097\/a742e326","article-title":"Research on pruning methods for mobileNet convolutional neural network","volume":"81","author":"Zou","year":"2024","journal-title":"Highlights Sci. Eng. Technol"},{"key":"B65","first-page":"392","article-title":"\u201cSPot-the-difference self-supervised pre-training for anomaly detection and segmentation,\u201d","volume-title":"Lecture Notes in Computer Science","author":"Zou","year":"2022"}],"container-title":["Frontiers in Computer Science"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcomp.2025.1563942\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,4,17]],"date-time":"2025-04-17T05:24:14Z","timestamp":1744867454000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcomp.2025.1563942\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,4,17]]},"references-count":65,"alternative-id":["10.3389\/fcomp.2025.1563942"],"URL":"https:\/\/doi.org\/10.3389\/fcomp.2025.1563942","relation":{},"ISSN":["2624-9898"],"issn-type":[{"value":"2624-9898","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,4,17]]},"article-number":"1563942"}}