{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,14]],"date-time":"2026-07-14T14:56:33Z","timestamp":1784040993815,"version":"3.55.0"},"reference-count":60,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"12","license":[{"start":{"date-parts":[[2020,12,1]],"date-time":"2020-12-01T00:00:00Z","timestamp":1606780800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2020,12,1]],"date-time":"2020-12-01T00:00:00Z","timestamp":1606780800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2020,12,1]],"date-time":"2020-12-01T00:00:00Z","timestamp":1606780800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100000923","name":"Australian Research Council","doi-asserted-by":"publisher","award":["FL-170100117"],"award-info":[{"award-number":["FL-170100117"]}],"id":[{"id":"10.13039\/501100000923","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000923","name":"Australian Research Council","doi-asserted-by":"publisher","award":["DP-180103424"],"award-info":[{"award-number":["DP-180103424"]}],"id":[{"id":"10.13039\/501100000923","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Neural Netw. Learning Syst."],"published-print":{"date-parts":[[2020,12]]},"DOI":"10.1109\/tnnls.2020.2966319","type":"journal-article","created":{"date-parts":[[2020,2,5]],"date-time":"2020-02-05T22:36:29Z","timestamp":1580942189000},"page":"5349-5362","source":"Crossref","is-referenced-by-count":369,"title":["Why ResNet Works? Residuals Generalize"],"prefix":"10.1109","volume":"31","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-5584-2385","authenticated-orcid":false,"given":"Fengxiang","family":"He","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9640-6472","authenticated-orcid":false,"given":"Tongliang","family":"Liu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7225-5449","authenticated-orcid":false,"given":"Dacheng","family":"Tao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref39","author":"vapnik","year":"1974","journal-title":"Theory of Pattern Recognition"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.2018.2854560"},{"key":"ref33","first-page":"2343","article-title":"When and why are deep networks better than shallow ones?","author":"mhaskar","year":"2017","journal-title":"Proc AAAI Conf Artif Intell"},{"key":"ref32","article-title":"Depth with nonlinearity creates no bad local minima in ResNets","author":"kawaguchi","year":"2018","journal-title":"arXiv 1810 09038"},{"key":"ref31","first-page":"15660","article-title":"Are deep resnets provably better than linear predictors?","author":"yun","year":"2019","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref30","first-page":"507","article-title":"Are resnets provably better than linear predictors?","author":"shamir","year":"2018","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref37","article-title":"Understanding generalization in recurrent neural networks","author":"tu","year":"2020","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref36","first-page":"1141","article-title":"Control batch size and learning rate to generalize well: Theoretical and empirical evidence","author":"he","year":"2019","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref35","first-page":"1947","article-title":"Emergence of invariance and disentanglement in deep representations","volume":"19","author":"achille","year":"2018","journal-title":"J Mach Learn Res"},{"key":"ref34","article-title":"Opening the black box of deep neural networks via information","author":"shwartz-ziv","year":"2017","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref60","first-page":"1","article-title":"Remarques sur un résultat non publié de B. Maurey","author":"pisier","year":"1981","journal-title":"Séminaire Analyse Fonctionnelle"},{"key":"ref28","article-title":"A PAC-Bayesian approach to spectrally-normalized margin bounds for neural networks","author":"neyshabur","year":"2018","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref27","article-title":"Sensitivity and generalization in neural networks: An empirical study","author":"novak","year":"2018","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref29","first-page":"6391","article-title":"Visualizing the loss landscape of neural nets","author":"li","year":"2018","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2016.2582924"},{"key":"ref1","doi-asserted-by":"crossref","first-page":"436","DOI":"10.1038\/nature14539","article-title":"Deep learning","volume":"521","author":"lecun","year":"2015","journal-title":"Nature"},{"key":"ref20","article-title":"Understanding deep learning requires rethinking generalization","author":"zhang","year":"2017","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref22","first-page":"1064","article-title":"Nearly-tight VC-dimension bounds for piecewise linear neural networks","author":"harvey","year":"2017","journal-title":"Proc Annu Conf Learn Theory"},{"key":"ref21","article-title":"Generalization in deep learning","author":"kawaguchi","year":"2017","journal-title":"arXiv 1710 05468"},{"key":"ref24","first-page":"297","article-title":"Size-independent sample complexity of neural networks","author":"golowich","year":"2018","journal-title":"Proc Annu Conf Learn Theory"},{"key":"ref23","author":"mohri","year":"2012","journal-title":"Foundations of Machine Learning"},{"key":"ref26","article-title":"Fisher-Rao metric, geometry, and complexity of neural networks","author":"liang","year":"2017","journal-title":"arXiv 1711 01530"},{"key":"ref25","first-page":"5947","article-title":"Exploring generalization in deep learning","author":"neyshabur","year":"2017","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref50","first-page":"527","article-title":"Covering number bounds of certain regularized linear function classes","volume":"2","author":"zhang","year":"2002","journal-title":"J Mach Learn Res"},{"key":"ref51","first-page":"1225","article-title":"Statistical analysis of some multi-category large margin classification methods","volume":"5","author":"zhang","year":"2004","journal-title":"J Mach Learn Res"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2018.2886995"},{"key":"ref58","article-title":"Closing the generalization gap of adaptive gradient methods in training deep neural networks","author":"chen","year":"2018","journal-title":"arXiv 1806 06763"},{"key":"ref57","article-title":"Three mechanisms of weight decay regularization","author":"zhang","year":"2019","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref56","article-title":"Adversarial training versus weight decay","author":"galloway","year":"2018","journal-title":"arXiv 1804 03308"},{"key":"ref55","first-page":"1019","article-title":"Sharp minima can generalize for deep nets","author":"dinh","year":"2017","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref54","article-title":"On large-batch training for deep learning: Generalization gap and sharp minima","author":"keskar","year":"2017","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref53","article-title":"Learning multiple layers of features from tiny images","author":"krizhevsky","year":"2009"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4419-5821-1_19"},{"key":"ref10","first-page":"1097","article-title":"ImageNet classification with deep convolutional neural networks","author":"krizhevsky","year":"2012","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref11","first-page":"265","article-title":"TensorFlow: A system for large-scale machine learning","volume":"16","author":"abadi","year":"2016","journal-title":"Proc USENIX Symp on Operating System Design and Implementation"},{"key":"ref40","first-page":"463","article-title":"Rademacher and Gaussian complexities: Risk bounds and structural results","volume":"3","author":"bartlett","year":"2002","journal-title":"J Mach Learn Res"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.106"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.322"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2017.2699184"},{"key":"ref15","author":"witten","year":"2016","journal-title":"Data Mining Practical Machine Learning Tools and Techniques"},{"key":"ref16","article-title":"Computer vision for autonomous vehicles: Problems, datasets and state of the art","author":"janai","year":"2017","journal-title":"arXiv 1704 05519"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1038\/nature21056"},{"key":"ref18","first-page":"950","article-title":"A simple weight decay can improve generalization","author":"krogh","year":"1992","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref19","first-page":"6240","article-title":"Spectrally-normalized margin bounds for neural networks","author":"bartlett","year":"2017","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1038\/nature16961"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2016.2618340"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2017.2764960"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2017.07.005"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.243"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2018.2873649"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.634"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2017.2690683"},{"key":"ref45","first-page":"2159","article-title":"Algorithmic stability and hypothesis complexity","author":"liu","year":"2017","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref48","first-page":"2336","article-title":"Generalization error bounds for optimization algorithms via stability","author":"meng","year":"2017","journal-title":"Proc AAAI Conf Artif Intell"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2017.2785329"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1016\/0097-3165(95)90052-7"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4419-5821-1_11"},{"key":"ref44","first-page":"196","article-title":"Algorithmic stability and generalization performance","author":"bousquet","year":"2001","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1214\/009053605000000282"}],"container-title":["IEEE Transactions on Neural Networks and Learning Systems"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/5962385\/9273274\/08984747.pdf?arnumber=8984747","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,4,27]],"date-time":"2022-04-27T17:19:37Z","timestamp":1651079977000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/8984747\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,12]]},"references-count":60,"journal-issue":{"issue":"12"},"URL":"https:\/\/doi.org\/10.1109\/tnnls.2020.2966319","relation":{},"ISSN":["2162-237X","2162-2388"],"issn-type":[{"value":"2162-237X","type":"print"},{"value":"2162-2388","type":"electronic"}],"subject":[],"published":{"date-parts":[[2020,12]]}}}