{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,29]],"date-time":"2026-07-29T14:52:54Z","timestamp":1785336774328,"version":"3.55.0"},"reference-count":31,"publisher":"Wiley","issue":"5","license":[{"start":{"date-parts":[[2026,2,25]],"date-time":"2026-02-25T00:00:00Z","timestamp":1771977600000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/onlinelibrary.wiley.com\/termsAndConditions#vor"},{"start":{"date-parts":[[2026,2,25]],"date-time":"2026-02-25T00:00:00Z","timestamp":1771977600000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62202123"],"award-info":[{"award-number":["62202123"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Concurrency and Computation"],"published-print":{"date-parts":[[2026,3]]},"abstract":"<jats:title>ABSTRACT<\/jats:title>\n                  <jats:p>\n                    With the widespread application of reinforcement learning and deep learning on edge devices, training neural networks on ARM architecture processors has become an urgent demand. However, existing mainstream deep learning frameworks are not sufficiently optimized for training workloads on ARM CPUs, resulting in low training efficiency. To address this problem, this paper proposes and implements a multi\u2010level AI training acceleration scheme based on the open\u2010source C++ library mlpack, targeting the slow training speed and high resource consumption of Convolutional Neural Networks (CNNs) on ARM platforms. The scheme accelerates training by deeply optimizing the im2col algorithm to convert convolutions into efficient matrix multiplications, utilizing ARM NEON SIMD instructions to optimize linear operators, and integrating an FP64\/FP16 mixed\u2010precision training strategy with dynamic loss scaling. Experimental results on LeNet\u20105 and VGG11\u2010style CNNs show substantial performance gains over the original mlpack and mainstream frameworks. On an NVIDIA Jetson AGX Orin, our implementation achieves up to\n                    <jats:bold>7.3<\/jats:bold>\n                    speedup over the original mlpack baseline and up to\n                    <jats:bold>11.3<\/jats:bold>\n                    \u00d7 and\n                    <jats:bold>5.69<\/jats:bold>\n                    \u00d7 end\u2010to\u2010end speedups over PyTorch and TensorFlow, respectively, while still delivering multi\u2010fold reductions in training time on a low\u2010resource Raspberry Pi platform. In DQN\u2010based reinforcement learning for Atari Breakout, our solution attains a\n                    <jats:bold>4.98<\/jats:bold>\n                    \u00d7 end\u2010to\u2010end speedup over the PyTorch single\u2010threaded baseline and maintains\n                    <jats:bold>2.37<\/jats:bold>\n                    \u00d7 and\n                    <jats:bold>4.23<\/jats:bold>\n                    \u00d7 advantages over 4\u2010threaded PyTorch and TensorFlow implementations. Ablation studies confirm the complementary nature of the proposed optimizations, with convolutional, linear, and mixed\u2010precision components jointly contributing to the overall speedup and enabling an attractive performance\u2013accuracy trade\u2010off for ARM\u2010based edge computing.\n                  <\/jats:p>","DOI":"10.1002\/cpe.70626","type":"journal-article","created":{"date-parts":[[2026,3,8]],"date-time":"2026-03-08T14:24:47Z","timestamp":1772979887000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":1,"title":["A Multi\u2010Level Acceleration Scheme for AI Model Training on ARM Architecture Processors"],"prefix":"10.1002","volume":"38","author":[{"ORCID":"https:\/\/orcid.org\/0009-0000-3422-8634","authenticated-orcid":false,"given":"Yang","family":"Bai","sequence":"first","affiliation":[{"name":"School of Cyberspace Science Harbin Institute of Technology  Harbin China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-4957-5672","authenticated-orcid":false,"given":"MingDong","family":"Xie","sequence":"additional","affiliation":[{"name":"School of Cyberspace Science Harbin Institute of Technology  Harbin China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Meng","family":"Hao","sequence":"additional","affiliation":[{"name":"School of Cyberspace Science Harbin Institute of Technology  Harbin China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4783-876X","authenticated-orcid":false,"given":"WeiZhe","family":"Zhang","sequence":"additional","affiliation":[{"name":"School of Cyberspace Science Harbin Institute of Technology  Harbin China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"HongWei","family":"Yang","sequence":"additional","affiliation":[{"name":"School of Cyberspace Science Harbin Institute of Technology  Harbin China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Li","family":"He","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Mobile Network and Mobile Multimedia Technology, ZTE  Shenzhen China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Wei","family":"Duan","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Mobile Network and Mobile Multimedia Technology, ZTE  Shenzhen China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"NingCheng","family":"Wang","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Mobile Network and Mobile Multimedia Technology, ZTE  Shenzhen China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2026,2,25]]},"reference":[{"key":"e_1_2_13_2_1","doi-asserted-by":"publisher","DOI":"10.1109\/5.726791"},{"key":"e_1_2_13_3_1","article-title":"Imagenet Classification With Deep Convolutional Neural Networks","volume":"25","author":"Krizhevsky A.","year":"2012","journal-title":"Advances in Neural Information Processing Systems"},{"issue":"2","key":"e_1_2_13_4_1","article-title":"Intelligent 6G Wireless Network With Multi\u2010Dimensional Information Perception","volume":"21","author":"Yang B.","year":"2023","journal-title":"ZTE Communications"},{"issue":"2","key":"e_1_2_13_5_1","first-page":"1","article-title":"Editorial: Special Topic on Edge Intelligence for Internet of Things","volume":"19","author":"Liu C. H.","year":"2021","journal-title":"ZTE Communications"},{"key":"e_1_2_13_6_1","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2017.2761740"},{"key":"e_1_2_13_7_1","unstructured":"K.Yang Y.Zhou Z.Yang andY.Shi \u201cCommunication\u2010Efficient Edge Ai Inference Over Wireless Networks \u201d(2020). arXiv preprint arXiv:2004.13351."},{"key":"e_1_2_13_8_1","unstructured":"S.Han H.Mao andW. J.Dally \u201cDeep Compression: Compressing Deep Neural Networks With Pruning Trained Quantization and Huffman Coding \u201d(2015). arXiv preprint arXiv:1510.00149."},{"key":"e_1_2_13_9_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00286"},{"key":"e_1_2_13_10_1","doi-asserted-by":"publisher","DOI":"10.3390\/s21092984"},{"key":"e_1_2_13_11_1","first-page":"365","volume-title":"Proceedings of the European Conference on Computer Vision (ECCV)","author":"Zhang D.","year":"2018"},{"key":"e_1_2_13_12_1","unstructured":"A. G.Howard M.Zhu B.Chen et al. \u201cMobilenets: Efficient Convolutional Neural Networks for Mobile Vision Applications \u201d(2017). arXiv preprint arXiv:1704.04861."},{"key":"e_1_2_13_13_1","unstructured":"F. N.Iandola S.Han M. W.Moskewicz K.Ashraf W. J.Dally andK.Keutzer \u201cSqueezeNet: AlexNet\u2010Level Accuracy With 50x Fewer Parameters and < 0.5 MB Model Size \u201d(2016). arXiv preprint arXiv:1602.07360."},{"key":"e_1_2_13_14_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01264-9_8"},{"key":"e_1_2_13_15_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.01099"},{"key":"e_1_2_13_16_1","doi-asserted-by":"publisher","DOI":"10.1109\/TIV.2023.3318070"},{"key":"e_1_2_13_17_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00140"},{"key":"e_1_2_13_18_1","first-page":"525","volume-title":"Xnor\u2010net: Imagenet Classification Using Binary Convolutional Neural Networks","author":"Rastegari M.","year":"2016"},{"key":"e_1_2_13_19_1","unstructured":"S.Zhou Y.Wu Z.Ni X.Zhou H.Wen andY.Zou \u201cDorefa\u2010Net: Training Low Bitwidth Convolutional Neural Networks With Low Bitwidth Gradients \u201d(2016). arXiv preprint arXiv:1606.06160."},{"issue":"2","key":"e_1_2_13_20_1","first-page":"2","article-title":"Enabling Intelligence at Network Edge: An Overview of Federated Learning","volume":"18","author":"Yang H.","year":"2020","journal-title":"ZTE Communications"},{"key":"e_1_2_13_21_1","article-title":"Pytorch: An Imperative Style, High\u2010Performance Deep Learning Library","volume":"32","author":"Paszke A.","year":"2019","journal-title":"Advances in Neural Information Processing Systems"},{"issue":"1","key":"e_1_2_13_22_1","first-page":"801","article-title":"MLPACK: A Scalable C++ Machine Learning Library","volume":"14","author":"Curtin R. R.","year":"2013","journal-title":"Journal of Machine Learning Research"},{"key":"e_1_2_13_23_1","unstructured":"P.Micikevicius S.Narang J.Alben et al. \u201cMixed Precision Training \u201d(2017). arXiv preprint arXiv:1710.03740."},{"key":"e_1_2_13_24_1","unstructured":"M.Abadi A.Agarwal P.Barham et al. \u201cTensorflow: Large\u2010Scale Machine Learning on Heterogeneous Distributed Systems \u201d(2016). arXiv preprint arXiv:1603.04467."},{"key":"e_1_2_13_25_1","unstructured":"C.Sanderson \u201cArmadillo: An Open Source C++ Linear Algebra Library for Fast Prototyping and Computationally Intensive Experiments \u201d(2010)."},{"key":"e_1_2_13_26_1","unstructured":"R.Krishnamoorthi \u201cQuantizing Deep Convolutional Networks for Efficient Inference: A Whitepaper \u201d(2018). arXiv preprint arXiv:1806.08342."},{"key":"e_1_2_13_27_1","doi-asserted-by":"publisher","DOI":"10.1145\/3487553.3524856"},{"key":"e_1_2_13_28_1","doi-asserted-by":"publisher","DOI":"10.1145\/3458817.3476217"},{"key":"e_1_2_13_29_1","doi-asserted-by":"publisher","DOI":"10.1145\/3650200.3656620"},{"key":"e_1_2_13_30_1","doi-asserted-by":"publisher","DOI":"10.1145\/3392717.3392765"},{"key":"e_1_2_13_31_1","doi-asserted-by":"publisher","DOI":"10.1007\/s41870-023-01320-9"},{"key":"e_1_2_13_32_1","doi-asserted-by":"publisher","DOI":"10.1002\/cpe.6102"}],"container-title":["Concurrency and Computation: Practice and Experience"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/cpe.70626","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/full-xml\/10.1002\/cpe.70626","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/cpe.70626","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,9]],"date-time":"2026-03-09T00:13:43Z","timestamp":1773015223000},"score":1,"resource":{"primary":{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/10.1002\/cpe.70626"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,25]]},"references-count":31,"journal-issue":{"issue":"5","published-print":{"date-parts":[[2026,3]]}},"alternative-id":["10.1002\/cpe.70626"],"URL":"https:\/\/doi.org\/10.1002\/cpe.70626","archive":["Portico"],"relation":{},"ISSN":["1532-0626","1532-0634"],"issn-type":[{"value":"1532-0626","type":"print"},{"value":"1532-0634","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,25]]},"assertion":[{"value":"2025-11-18","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-02-09","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-02-25","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70626"}}