{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,4]],"date-time":"2026-07-04T17:10:46Z","timestamp":1783185046260,"version":"3.54.6"},"reference-count":56,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","license":[{"start":{"date-parts":[[2022,1,1]],"date-time":"2022-01-01T00:00:00Z","timestamp":1640995200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2022,1,1]],"date-time":"2022-01-01T00:00:00Z","timestamp":1640995200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2022,1,1]],"date-time":"2022-01-01T00:00:00Z","timestamp":1640995200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["61977046, 62171088"],"award-info":[{"award-number":["61977046, 62171088"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100003399","name":"Science and Technology Commission of Shanghai Municipality","doi-asserted-by":"publisher","award":["2021SHZDZX0102"],"award-info":[{"award-number":["2021SHZDZX0102"]}],"id":[{"id":"10.13039\/501100003399","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Pattern Anal. Mach. Intell."],"published-print":{"date-parts":[[2022]]},"DOI":"10.1109\/tpami.2022.3178101","type":"journal-article","created":{"date-parts":[[2022,5,26]],"date-time":"2022-05-26T19:39:27Z","timestamp":1653593967000},"page":"1-1","source":"Crossref","is-referenced-by-count":15,"title":["Low Dimensional Trajectory Hypothesis is True: DNNs can be Trained in Tiny Subspaces"],"prefix":"10.1109","author":[{"given":"Tao","family":"Li","sequence":"first","affiliation":[{"name":"Department of Automation, Shanghai Jiao Tong Univeristy, Shanghai, 200240, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lei","family":"Tan","sequence":"additional","affiliation":[{"name":"Department of Automation, Shanghai Jiao Tong Univeristy, Shanghai, 200240, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhehao","family":"Huang","sequence":"additional","affiliation":[{"name":"Department of Automation, Shanghai Jiao Tong Univeristy, Shanghai, 200240, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qinghua","family":"Tao","sequence":"additional","affiliation":[{"name":"Department of Automation, Tsinghua University, Beijing, 100084, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yipeng","family":"Liu","sequence":"additional","affiliation":[{"name":"School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaolin","family":"Huang","sequence":"additional","affiliation":[{"name":"Department of Automation, Shanghai Jiao Tong University, 12474 Shanghai, Shanghai, China, 200240"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/5.726791"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1038\/nature14539"},{"key":"ref3","article-title":"Very deep convolutional networks for large-scale image recognition","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Simonyan"},{"key":"ref4","article-title":"MobileNets: Efficient convolutional neural networks for mobile vision applications","author":"Howard","year":"2017"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.195"},{"key":"ref6","article-title":"Gradient descent happens in a tiny subspace","author":"Gur-Ari","year":"2018"},{"key":"ref7","article-title":"Measuring the intrinsic dimension of objective landscapes","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Li"},{"key":"ref8","article-title":"Learning multiple layers of features from tiny images","author":"Krizhevsky","year":"2009"},{"key":"ref9","article-title":"LeNet-5, convolutional neural networks","author":"LeCun","year":"2015"},{"key":"ref10","first-page":"12140","article-title":"Improving neural network training in low dimensional random bases","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Gressmann"},{"key":"ref11","first-page":"8571","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Jacot"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1088\/1742-5468\/abc62b"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.5244\/C.29.31"},{"key":"ref14","article-title":"The lottery ticket hypothesis: Finding sparse, trainable neural networks","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Frankle"},{"key":"ref15","article-title":"An overview of gradient descent optimization algorithms","author":"Ruder","year":"2016"},{"key":"ref16","article-title":"Adam: A method for stochastic optimization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Kingma"},{"key":"ref17","article-title":"Equilibrated adaptive learning rates for non-convex optimization","author":"Dauphin","year":"2015"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1093\/imamat\/6.1.76"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1007\/BF01582063"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1145\/3446776"},{"key":"ref21","first-page":"6389","article-title":"Visualizing the loss landscape of neural nets","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Li"},{"key":"ref22","first-page":"2549","article-title":"Asymmetric valleys: Beyond sharp and flat local minima","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"He"},{"key":"ref23","article-title":"Quasi-newtons method in the class gradient defined high-curvature subspace","author":"Tuddenham","year":"2020"},{"key":"ref24","first-page":"604","article-title":"Fast large-scale optimization by unifying stochastic gradient and quasi-newton methods","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Sohl-Dickstein"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2013.6638949"},{"key":"ref26","article-title":"Initialization and regularization of factorized neural layers","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Khodak"},{"key":"ref27","first-page":"2790","article-title":"Parameter-efficient transfer learning for NLP","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Houlsby"},{"key":"ref28","article-title":"LoRA: Low-rank adaptation of large language models","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Hu"},{"key":"ref29","first-page":"7710","article-title":"Spectra of the conjugate kernel and neural tangent kernel for linear-width neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Fan"},{"key":"ref30","first-page":"1451","article-title":"On the similarity between the laplace and neural tangent kernels","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Geifman"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref32","first-page":"244","article-title":"Adaptive bound optimization for online convex optimization","volume-title":"Proc. Conf. Learn. Theory","author":"McMahan"},{"issue":"7","key":"ref33","first-page":"2121","article-title":"Adaptive subgradient methods for online learning and stochastic optimization","volume":"12","author":"Duchi","year":"2011","journal-title":"J. Mach. Learn. Res."},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1007\/BF01589116"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1093\/imanum\/11.3.325"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/TSP.2014.2357775"},{"key":"ref37","first-page":"557","article-title":"Practical Gauss-Newton optimisation for deep learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Botev"},{"key":"ref38","first-page":"620","article-title":"A progressive batching L-BFGS method for machine learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Bollapragada"},{"key":"ref39","first-page":"2386","article-title":"Practical quasi-Newton methods for training deep neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Goldfarb"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.2140\/pjm.1966.16.1"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.243"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.308"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00907"},{"key":"ref45","first-page":"6105","article-title":"EfficientNet: Rethinking model scaling for convolutional neural networks","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Tan"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.4324\/9781410605337-29"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01264-9_8"},{"key":"ref48","article-title":"SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size","author":"Iandola","year":"2016"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00745"},{"key":"ref50","first-page":"4313","article-title":"Gradient descent with early stopping is provably robust to label noise for overparameterized neural networks","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Li"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v31i1.10894"},{"key":"ref52","first-page":"8792","article-title":"Generalized cross entropy loss for training deep neural networks with noisy labels","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhang"},{"key":"ref53","article-title":"Averaging weights leads to wider optima and better generalization","author":"Izmailov","year":"2018"},{"key":"ref54","first-page":"242","article-title":"A convergence theory for deep learning via over-parameterization","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Allen-Zhu"},{"key":"ref55","article-title":"Prototypical networks for few-shot learning","author":"Snell","year":"2017"},{"issue":"2","key":"ref56","doi-asserted-by":"crossref","first-page":"77","DOI":"10.1023\/A:1019956318069","article-title":"A perspective view and survey of meta-learning","volume":"18","author":"Vilalta","year":"2002","journal-title":"Artif. Intell. Rev."}],"container-title":["IEEE Transactions on Pattern Analysis and Machine Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/34\/4359286\/09782552.pdf?arnumber=9782552","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,2,1]],"date-time":"2024-02-01T01:53:27Z","timestamp":1706752407000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9782552\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2022]]},"references-count":56,"URL":"https:\/\/doi.org\/10.1109\/tpami.2022.3178101","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":[[2022]]}}}