{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,8,15]],"date-time":"2025-08-15T01:06:36Z","timestamp":1755219996388,"version":"3.43.0"},"reference-count":80,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"9","license":[{"start":{"date-parts":[[2025,9,1]],"date-time":"2025-09-01T00:00:00Z","timestamp":1756684800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,9,1]],"date-time":"2025-09-01T00:00:00Z","timestamp":1756684800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,9,1]],"date-time":"2025-09-01T00:00:00Z","timestamp":1756684800000},"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":["62025208","62421002"],"award-info":[{"award-number":["62025208","62421002"]}],"id":[{"id":"10.13039\/501100001809","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":[[2025,9]]},"DOI":"10.1109\/tpami.2025.3572251","type":"journal-article","created":{"date-parts":[[2025,5,21]],"date-time":"2025-05-21T13:40:05Z","timestamp":1747834805000},"page":"7976-7986","source":"Crossref","is-referenced-by-count":0,"title":["Toward Understanding the Generalizability of Delayed Stochastic Gradient Descent"],"prefix":"10.1109","volume":"47","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0622-1202","authenticated-orcid":false,"given":"Xiaoge","family":"Deng","sequence":"first","affiliation":[{"name":"Intelligent Game and Decision Lab, Beijing, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5659-3464","authenticated-orcid":false,"given":"Li","family":"Shen","sequence":"additional","affiliation":[{"name":"Sun Yat-sen University, Guangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7419-1511","authenticated-orcid":false,"given":"Shengwei","family":"Li","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5024-1900","authenticated-orcid":false,"given":"Tao","family":"Sun","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9743-2034","authenticated-orcid":false,"given":"Dongsheng","family":"Li","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7225-5449","authenticated-orcid":false,"given":"Dacheng","family":"Tao","sequence":"additional","affiliation":[{"name":"JD Explore Academy, Beijing, China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1214\/aoms\/1177729586"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1137\/16M1080173"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1145\/3446776"},{"article-title":"Gradient descent provably optimizes over-parameterized neural networks","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Du","key":"ref4"},{"key":"ref5","first-page":"1225","article-title":"Train faster, generalize better: Stability of stochastic gradient descent","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Hardt"},{"key":"ref6","first-page":"2815","article-title":"Data-dependent stability of stochastic gradient descent","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Kuzborskij"},{"key":"ref7","first-page":"5809","article-title":"Fine-grained analysis of stability and generalization for stochastic gradient descent","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Lei"},{"key":"ref8","first-page":"4381","article-title":"Stability of stochastic gradient descent on nonsmooth convex losses","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Bassily"},{"key":"ref9","first-page":"2364","article-title":"Stability of SGD: Tightness analysis and improved bounds","volume-title":"Proc. Int. Conf. Uncertainty Artif. Intell.","author":"Zhang"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1007\/s10994-021-06056-w"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"ref12","first-page":"1223","article-title":"Large scale distributed deep networks","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Dean"},{"key":"ref13","first-page":"1877","article-title":"Language models are few-shot learners","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Brown"},{"key":"ref14","first-page":"2595","article-title":"Parallelized stochastic gradient descent","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Zinkevich"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.5555\/2685048.2685095"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2020.3026619"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1016\/s1570-579x(01)80023-9"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/cdc.2012.6426626"},{"key":"ref19","first-page":"2737","article-title":"Asynchronous parallel stochastic gradient for nonconvex optimization","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Lian"},{"key":"ref20","first-page":"5970","article-title":"Distributed asynchronous optimization with unbounded delays: How slow can you go?","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Zhou"},{"key":"ref21","first-page":"111","article-title":"A tight convergence analysis for stochastic gradient descent with delayed updates","volume-title":"Proc. 31st Int. Conf. Algorithmic Learn. Theory","author":"Arjevani"},{"issue":"237","key":"ref22","first-page":"1","article-title":"The error-feedback framework: Better rates for SGD with delayed gradients and compressed updates","volume":"21","author":"Stich","year":"2020","journal-title":"J. Mach. Learn. Res."},{"key":"ref23","first-page":"9024","article-title":"Asynchronous stochastic optimization robust to arbitrary delays","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Cohen"},{"key":"ref24","first-page":"420","article-title":"Asynchronous SGD beats minibatch SGD under arbitrary delays","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Mishchenko"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/ALLERTON.2019.8919791"},{"key":"ref26","first-page":"6391","article-title":"Visualizing the loss landscape of neural nets","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Li"},{"key":"ref27","first-page":"8580","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Jacot"},{"key":"ref28","first-page":"3325","article-title":"The power of interpolation: Understanding the effectiveness of SGD in modern over-parametrized learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Ma"},{"key":"ref29","first-page":"1141","article-title":"Control batch size and learning rate to generalize well: Theoretical and empirical evidence","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"He"},{"key":"ref30","first-page":"4633","article-title":"Benign overfitting of constant-stepsize SGD for linear regression","volume-title":"Proc. 34th Conf. Learn. Theory","author":"Zou"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.1986.1104412"},{"key":"ref32","first-page":"693","article-title":"HOGWILD!: A lock-free approach to parallelizing stochastic gradient descent","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Recht"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1137\/16M1057000"},{"key":"ref34","first-page":"6182","article-title":"Asynchronous coordinate descent under more realistic assumptions","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Sun"},{"key":"ref35","first-page":"436","article-title":"Asynchronous distributed learning: Adapting to gradient delays without prior knowledge","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Aviv"},{"key":"ref36","first-page":"4120","article-title":"Asynchronous stochastic gradient descent with delay compensation","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Zheng"},{"key":"ref37","first-page":"957","article-title":"AdaDelay: Delay adaptive distributed stochastic optimization","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Sra"},{"key":"ref38","first-page":"2350","article-title":"Staleness-aware async-SGD for distributed deep learning","volume-title":"Proc. 25th Int. Joint Conf. Artif. Intell.","author":"Zhang"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i04.6001"},{"key":"ref40","first-page":"1261","article-title":"ASAP.SGD: Instance-based adaptiveness to staleness in asynchronous SGD","volume-title":"Proc. 39th Int. Conf. Mach. Learn.","author":"B\u00e4ckstr\u00f6m"},{"key":"ref41","first-page":"1928","article-title":"Asynchronous methods for deep reinforcement learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Mnih"},{"key":"ref42","first-page":"17202","article-title":"Sharper convergence guarantees for asynchronous SGD for distributed and federated learning","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Koloskova"},{"issue":"78","key":"ref43","first-page":"1","article-title":"Multi-agent online optimization with delays: Asynchronicity, adaptivity, and optimism","volume":"23","author":"Hsieh","year":"2022","journal-title":"J. Mach. Learn. Res."},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1214\/aos\/1176344196"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.1979.1056032"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.1979.1056018"},{"key":"ref47","first-page":"499","article-title":"Stability and generalization","volume":"2","author":"Bousquet","year":"2002","journal-title":"J. Mach. Learn. Res."},{"issue":"3","key":"ref48","first-page":"55","article-title":"Stability of randomized learning algorithms","volume":"6","author":"Elisseeff","year":"2005","journal-title":"J. Mach. Learn. Res."},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1007\/s10444-004-7634-z"},{"issue":"90","key":"ref50","first-page":"2635","article-title":"Learnability, stability and uniform convergence","volume":"11","author":"Shalev-Shwartz","year":"2010","journal-title":"J. Mach. Learn. Res."},{"key":"ref51","first-page":"745","article-title":"Stability and generalization of learning algorithms that converge to global optima","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Charles"},{"key":"ref52","first-page":"9770","article-title":"Generalization bounds for uniformly stable algorithms","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Feldman"},{"key":"ref53","first-page":"1270","article-title":"High probability generalization bounds for uniformly stable algorithms with nearly optimal rate","volume-title":"Proc. Int. Conf. Learn. Theory","author":"Feldman"},{"key":"ref54","first-page":"610","article-title":"Sharper bounds for uniformly stable algorithms","volume-title":"Proc. Int. Conf. Learn. Theory","author":"Bousquet"},{"key":"ref55","first-page":"34241","article-title":"On the generalization of learning algorithms that do not converge","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Chandramoorthy"},{"key":"ref56","first-page":"8609","article-title":"Stability & generalisation of gradient descent for shallow neural networks without the neural tangent kernel","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Richards"},{"key":"ref57","first-page":"38557","article-title":"Stability and generalization analysis of gradient methods for shallow neural networks","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Lei"},{"key":"ref58","first-page":"605","article-title":"Generalization bounds of SGLD for non-convex learning: Two theoretical viewpoints","volume-title":"Proc. Int. Conf. Learn. Theory","author":"Mou"},{"key":"ref59","first-page":"1412","article-title":"Stability based generalization bounds for exponential family Langevin dynamics","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Banerjee"},{"key":"ref60","first-page":"21236","article-title":"Sharper generalization bounds for pairwise learning","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Lei"},{"key":"ref61","first-page":"20160","article-title":"Simple stochastic and online gradient descent algorithms for pairwise learning","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Yang"},{"key":"ref62","first-page":"2173","article-title":"Generalization bounds for meta-learning via PAC-Bayes and uniform stability","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Farid"},{"key":"ref63","first-page":"18487","article-title":"Fine-grained analysis of stability and generalization for modern meta learning algorithms","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Guan"},{"key":"ref64","first-page":"3174","article-title":"Train simultaneously, generalize better: Stability of gradient-based minimax learners","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Farnia"},{"key":"ref65","first-page":"6175","article-title":"Stability and generalization of stochastic gradient methods for minimax problems","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Lei"},{"key":"ref66","first-page":"26523","article-title":"On the algorithmic stability of adversarial training","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Xing"},{"key":"ref67","first-page":"38190","article-title":"What is a good metric to study generalization of minimax learners?","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Ozdaglar"},{"key":"ref68","first-page":"15446","article-title":"Stability analysis and generalization bounds of adversarial training","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Xiao"},{"key":"ref69","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2019.2910188"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v35i11.17173"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v37i6.25894"},{"key":"ref72","first-page":"27479","article-title":"Topology-aware generalization of decentralized SGD","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Zhu"},{"key":"ref73","doi-asserted-by":"publisher","DOI":"10.1016\/B978-0-12-751955-5.50005-8"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1017\/CBO9780511609589"},{"key":"ref75","doi-asserted-by":"publisher","DOI":"10.1017\/CBO9780511801655.003"},{"key":"ref76","first-page":"8570","article-title":"Wide neural networks of any depth evolve as linear models under gradient descent","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Lee"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1145\/1961189.1961199"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.1109\/5.726791"},{"article-title":"Learning multiple layers of features from tiny images","year":"2009","author":"Krizhevsky","key":"ref80"}],"container-title":["IEEE Transactions on Pattern Analysis and Machine Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/34\/11118328\/11008843.pdf?arnumber=11008843","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,8,7]],"date-time":"2025-08-07T17:44:41Z","timestamp":1754588681000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11008843\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,9]]},"references-count":80,"journal-issue":{"issue":"9"},"URL":"https:\/\/doi.org\/10.1109\/tpami.2025.3572251","relation":{},"ISSN":["0162-8828","2160-9292","1939-3539"],"issn-type":[{"type":"print","value":"0162-8828"},{"type":"electronic","value":"2160-9292"},{"type":"electronic","value":"1939-3539"}],"subject":[],"published":{"date-parts":[[2025,9]]}}}