{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,8]],"date-time":"2026-04-08T16:48:37Z","timestamp":1775666917054,"version":"3.50.1"},"reference-count":55,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"6","license":[{"start":{"date-parts":[[2024,6,1]],"date-time":"2024-06-01T00:00:00Z","timestamp":1717200000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2024,6,1]],"date-time":"2024-06-01T00:00:00Z","timestamp":1717200000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2024,6,1]],"date-time":"2024-06-01T00:00:00Z","timestamp":1717200000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"National Key R&D Program of China","award":["2022YFA1008200"],"award-info":[{"award-number":["2022YFA1008200"]}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["92270001"],"award-info":[{"award-number":["92270001"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["12371511"],"award-info":[{"award-number":["12371511"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Shanghai Municipal of Science and Technology Major","award":["2021SHZDZX0102"],"award-info":[{"award-number":["2021SHZDZX0102"]}]},{"name":"HPC of School of Mathematical Sciences"},{"name":"Student Innovation Center"},{"name":"Siyuan-1 cluster"},{"name":"Center for High Performance Computing at Shanghai Jiao Tong University"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Pattern Anal. Mach. Intell."],"published-print":{"date-parts":[[2024,6]]},"DOI":"10.1109\/tpami.2024.3357172","type":"journal-article","created":{"date-parts":[[2024,1,23]],"date-time":"2024-01-23T20:47:39Z","timestamp":1706042859000},"page":"4206-4217","source":"Crossref","is-referenced-by-count":42,"title":["Implicit Regularization of Dropout"],"prefix":"10.1109","volume":"46","author":[{"ORCID":"https:\/\/orcid.org\/0009-0006-4202-8556","authenticated-orcid":false,"given":"Zhongwang","family":"Zhang","sequence":"first","affiliation":[{"name":"Institute of Natural Sciences, School of Mathematical Sciences, MOE-LSC and Qing Yuan Research Institute, Shanghai Jiao Tong University, Shanghai, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0627-3520","authenticated-orcid":false,"given":"Zhi-Qin John","family":"Xu","sequence":"additional","affiliation":[{"name":"Institute of Natural Sciences, School of Mathematical Sciences, MOE-LSC and Qing Yuan Research Institute, Shanghai Jiao Tong University, Shanghai, China"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Improving neural networks by preventing co-adaptation of feature detectors","author":"Hinton","year":"2012"},{"issue":"1","key":"ref2","first-page":"1929","article-title":"Dropout: A simple way to prevent neural networks from overfitting","volume":"15","author":"Srivastava","year":"2014","journal-title":"J. Mach. Learn. Res."},{"key":"ref3","first-page":"6105","article-title":"EfficientNet: Rethinking model scaling for convolutional neural networks","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Tan"},{"issue":"1","key":"ref4","first-page":"3403","article-title":"On the inductive bias of dropout","volume":"16","author":"Helmbold","year":"2015","journal-title":"J. Mach. Learn. Res."},{"key":"ref5","first-page":"3645","article-title":"Dropout training, data-dependent regularization, and generalization bounds","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Mou"},{"key":"ref6","first-page":"351","article-title":"Dropout: Explicit forms and capacity control","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Arora"},{"key":"ref7","first-page":"3540","article-title":"On the implicit bias of dropout","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Mianjy"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1017\/S0962492902000144"},{"issue":"71","key":"ref9","first-page":"1","article-title":"Phase diagram for two-layer ReLU neural networks at infinite-width limit","volume":"22","author":"Luo","year":"2021","journal-title":"J. Mach. Learn. Res."},{"key":"ref10","article-title":"Towards understanding the condensation of neural networks at initial training","author":"Zhou","year":"2021"},{"key":"ref11","first-page":"26021","article-title":"Empirical phase diagram for three-layer neural networks with infinite width","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhou"},{"key":"ref12","first-page":"8580","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Jacot"},{"key":"ref13","article-title":"On large-batch training for deep learning: Generalization gap and sharp minima","author":"Keskar","year":"2016"},{"key":"ref14","article-title":"Exploring generalization in deep learning","author":"Neyshabur","year":"2017"},{"key":"ref15","article-title":"The anisotropic noise in stochastic gradient descent: Its behavior of escaping from sharp minima and regularization effects","author":"Zhu","year":"2018"},{"key":"ref16","article-title":"A PAC-Bayesian tutorial with a dropout bound","author":"McAllester","year":"2013"},{"key":"ref17","first-page":"1058","article-title":"Regularization of neural networks using DropConnect","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Wan"},{"key":"ref18","article-title":"Adaptive dropout with rademacher complexity regularization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Zhai"},{"key":"ref19","first-page":"21151","article-title":"On convergence and generalization of dropout training","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Mianjy"},{"key":"ref20","first-page":"2814","article-title":"Understanding dropout","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Baldi"},{"key":"ref21","first-page":"351","article-title":"Dropout training as adaptive regularization","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Wager"},{"key":"ref22","first-page":"100","article-title":"Altitude training: Strong bounds for single-layer dropout","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Wager"},{"key":"ref23","article-title":"Almost sure convergence of dropout algorithms for neural networks","author":"Senen-Cerda","year":"2020"},{"key":"ref24","article-title":"Stochastic modified equations and dynamics of dropout algorithm","author":"Zhang","year":"2023"},{"key":"ref25","article-title":"An ETF view of dropout regularization","volume-title":"Proc. Brit. Mach. Vis. Conf.","author":"Bank"},{"key":"ref26","first-page":"7550","article-title":"Dropout as a regularizer of interaction effects","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Lengerich"},{"key":"ref27","first-page":"435","article-title":"Dropout as a low-rank regularizer for matrix factorization","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Cavazza"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00769"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW50498.2020.00124"},{"issue":"180","key":"ref30","first-page":"1","article-title":"Dropout training is distributionally robust optimal","volume":"24","author":"Blanchet","year":"2023","journal-title":"J. Mach. Learn. Res."},{"key":"ref31","first-page":"10181","article-title":"The implicit and explicit regularization effects of dropout","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Wei"},{"key":"ref32","article-title":"Implicit gradient regularization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Barrett"},{"key":"ref33","article-title":"On the origin of implicit regularization in stochastic gradient descent","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Smith"},{"key":"ref34","first-page":"8289","article-title":"How SGD selects the global minima in over-parameterized learning: A dynamical stability perspective","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Wu"},{"key":"ref35","article-title":"Fluctuation-dissipation relations for stochastic gradient descent","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Yaida"},{"key":"ref36","first-page":"14848","article-title":"Embedding principle of loss landscape of deep neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhang"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.4208\/jml.220108"},{"key":"ref38","article-title":"Gradient descent quantizes ReLU network features","author":"Maennel","year":"2018"},{"key":"ref39","first-page":"5356","article-title":"An analytic theory of shallow networks dynamics for hinge loss classification","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Pellegrini"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.4208\/csiam-am.so-2023-0020"},{"key":"ref41","article-title":"Linear stability hypothesis and rank stratification for nonlinear models","author":"Zhang","year":"2022"},{"key":"ref42","first-page":"483","article-title":"Implicit regularization for deep neural networks driven by an Ornstein-Uhlenbeck like process","volume-title":"Proc. Conf. Learn. Theory","author":"Blanc"},{"key":"ref43","article-title":"Visualizing the loss landscape of neural nets","author":"Li","year":"2017"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-36708-4_22"},{"key":"ref45","first-page":"5301","article-title":"On the spectral bias of deep neural networks","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Rahaman"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.4208\/cicp.OA-2020-0085"},{"key":"ref47","first-page":"4761","article-title":"The convergence rate of neural networks for learned functions of different frequencies","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Ronen"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.24963\/ijcai.2021\/304"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1088\/0256-307X\/38\/3\/038701"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.4208\/csiam-am.SO-2020-0005"},{"key":"ref51","article-title":"Overview frequency principle\/spectral bias in deep learning","author":"Xu","year":"2022"},{"key":"ref52","article-title":"Very deep convolutional networks for large-scale image recognition","author":"Simonyan","year":"2014"},{"key":"ref53","article-title":"Adam: A method for stochastic optimization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Kingma"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"}],"container-title":["IEEE Transactions on Pattern Analysis and Machine Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/34\/10522060\/10412142.pdf?arnumber=10412142","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,5,8]],"date-time":"2024-05-08T04:43:13Z","timestamp":1715143393000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10412142\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,6]]},"references-count":55,"journal-issue":{"issue":"6"},"URL":"https:\/\/doi.org\/10.1109\/tpami.2024.3357172","relation":{},"ISSN":["0162-8828","2160-9292","1939-3539"],"issn-type":[{"value":"0162-8828","type":"print"},{"value":"2160-9292","type":"electronic"},{"value":"1939-3539","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,6]]}}}