{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,10]],"date-time":"2026-03-10T19:28:52Z","timestamp":1773170932239,"version":"3.50.1"},"reference-count":71,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"4","license":[{"start":{"date-parts":[[2026,4,1]],"date-time":"2026-04-01T00:00:00Z","timestamp":1775001600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,4,1]],"date-time":"2026-04-01T00:00:00Z","timestamp":1775001600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,4,1]],"date-time":"2026-04-01T00:00:00Z","timestamp":1775001600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"NTU RSR","award":["Start Up Grants"],"award-info":[{"award-number":["Start Up Grants"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Pattern Anal. Mach. Intell."],"published-print":{"date-parts":[[2026,4]]},"DOI":"10.1109\/tpami.2025.3647762","type":"journal-article","created":{"date-parts":[[2025,12,24]],"date-time":"2025-12-24T18:45:42Z","timestamp":1766601942000},"page":"4744-4755","source":"Crossref","is-referenced-by-count":0,"title":["Toward Understanding Generalization and Stability Gaps Between Centralized and Decentralized Federated Learning"],"prefix":"10.1109","volume":"48","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-2271-252X","authenticated-orcid":false,"given":"Yan","family":"Sun","sequence":"first","affiliation":[{"name":"University of Sydney, Camperdown, Australia"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5659-3464","authenticated-orcid":false,"given":"Li","family":"Shen","sequence":"additional","affiliation":[{"name":"Shenzhen Campus of Sun Yat-sen University, Shenzhen, 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":"Nanyang Technological University, Singapore"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","first-page":"1273","article-title":"Communication-efficient learning of deep networks from decentralized data","volume-title":"Proc. Artif. Intell. Statist.","author":"McMahan","year":"2017"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2022.3196503"},{"key":"ref3","first-page":"5330","article-title":"Can decentralized algorithms outperform centralized algorithms? a case study for decentralized parallel stochastic gradient descent","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"30","author":"Lian","year":"2017"},{"key":"ref4","first-page":"31269","article-title":"Improving the model consistency of decentralized federated learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Shi","year":"2023"},{"issue":"1","key":"ref5","first-page":"55","article-title":"Stability of randomized learning algorithms","volume":"6","author":"Elisseeff","year":"2005","journal-title":"J. Mach. Learn. Res."},{"key":"ref6","first-page":"1225","article-title":"Train faster, generalize better: Stability of stochastic gradient descent","volume-title":"Proc. Int. Conf. Mach. Learn","author":"Hardt","year":"2016"},{"key":"ref7","article-title":"Achieving linear speedup with partial worker participation in Non-IID federated learning","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Yang","year":"2021"},{"key":"ref8","first-page":"5132","article-title":"Scaffold: Stochastic controlled averaging for federated learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Karimireddy","year":"2020"},{"key":"ref9","first-page":"429","article-title":"Federated optimization in heterogeneous networks","volume-title":"Proc. Mach. Learn. Syst.","volume":"2","author":"Li","year":"2020"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/TSP.2021.3115952"},{"key":"ref11","article-title":"Federated learning based on dynamic regularization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Acar","year":"2021"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/ICDE53745.2022.00238"},{"key":"ref13","article-title":"FedSpeed: Larger local interval, less communication round, and higher generalization accuracy","volume-title":"Proc. 11th Int. Conf. Learn. Representations","author":"Sun","year":"2023"},{"key":"ref14","article-title":"What do we mean by generalization in federated learning?","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Yuan","year":"2022"},{"key":"ref15","first-page":"21554","article-title":"Robust federated learning: The case of affine distribution shifts","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"33","author":"Reisizadeh","year":"2020"},{"key":"ref16","first-page":"18250","article-title":"Generalized federated learning via sharpness aware minimization","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Qu","year":"2022"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20050-2_38"},{"key":"ref18","first-page":"32991","article-title":"Dynamic regularized sharpness aware minimization in federated learning: Approaching global consistency and smooth landscape","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Sun","year":"2023"},{"key":"ref19","first-page":"44093","article-title":"Federated learning you may communicate less often!","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Sefidgaran","year":"2024"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/SPAWC48557.2020.9154277"},{"key":"ref21","first-page":"676","article-title":"Understanding generalization of federated learning via stability: Heterogeneity matters","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Sun","year":"2024"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3530"},{"key":"ref23","first-page":"3478","article-title":"Decentralized stochastic optimization and gossip algorithms with compressed communication","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Koloskova","year":"2019"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2022.3230938"},{"issue":"213","key":"ref25","first-page":"1","article-title":"Cooperative SGD: A unified framework for the design and analysis of local-update SGD algorithms","volume-title":"J. Mach. Learn. Res.","volume":"22","author":"Wang","year":"2021"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/TSP.2020.3008605"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1137\/14096668X"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/TSP.2022.3184770"},{"key":"ref29","first-page":"15750","article-title":"Proxskip: Yes! Local gradient steps provably lead to communication acceleration! finally!","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Mishchenko","year":"2022"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/CDC49753.2023.10383873"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2024.3383271"},{"key":"ref32","article-title":"Fully decentralized federated learning","volume-title":"Proc. 3rd Workshop Bayesian Deep Learn.","volume":"2","author":"Lalitha","year":"2018"},{"key":"ref33","first-page":"3043","article-title":"Asynchronous decentralized parallel stochastic gradient descent","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Lian","year":"2018"},{"key":"ref34","first-page":"7184","article-title":"On the linear speedup analysis of communication efficient momentum SGD for distributed non-convex optimization","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Yu","year":"2019"},{"key":"ref35","first-page":"344","article-title":"Stochastic gradient push for distributed deep learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Assran","year":"2019"},{"key":"ref36","first-page":"5381","article-title":"A unified theory of decentralized SGD with changing topology and local updates","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Koloskova","year":"2020"},{"key":"ref37","article-title":"Decentralized federated learning: A segmented gossip approach","author":"Hu","year":"2019"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1016\/j.jpdc.2020.10.006"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2021.3096076"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/CCNC49033.2022.9700624"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2021.3138977"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2025.3546659"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v35i11.17173"},{"key":"ref44","first-page":"43005","article-title":"Decentralized SGD and average-direction SAM are asymptotically equivalent","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Zhu","year":"2023"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1561\/0400000003"},{"key":"ref46","first-page":"27290","article-title":"Towards understanding why lookahead generalizes better than SGD and beyond","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"34","author":"Zhou","year":"2021"},{"key":"ref47","article-title":"Adaptive federated optimization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Reddi","year":"2021"},{"key":"ref48","first-page":"3556","article-title":"Local SGD: Unified theory and new efficient methods","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Gorbunov","year":"2021"},{"key":"ref49","article-title":"FEDCM: Federated learning with client-level momentum","author":"Xu","year":"2021"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2023.3300886"},{"key":"ref51","article-title":"Fusion of global and local knowledge for personalized federated learning","author":"Huang","year":"2023","journal-title":"Trans. Mach. Learn. Res."},{"key":"ref52","first-page":"15446","article-title":"Stability analysis and generalization bounds of adversarial training","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"35","author":"Xiao","year":"2022"},{"key":"ref53","first-page":"27479","article-title":"Topology-aware generalization of decentralized SGD","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Zhu","year":"2022"},{"key":"ref54","first-page":"5562","article-title":"The lipschitz constant of self-attention","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Kim","year":"2021"},{"key":"ref55","first-page":"7325","article-title":"Stability and convergence of stochastic gradient clipping: Beyond Lipschitz continuity and smoothness","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Mai","year":"2021"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1137\/22m1527210"},{"key":"ref57","first-page":"7066","article-title":"Beyond uniform lipschitz condition in differentially private optimization","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Das","year":"2023"},{"key":"ref58","first-page":"20461","article-title":"On large-cohort training for federated learning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"34","author":"Charles","year":"2021"},{"key":"ref59","article-title":"On the convergence of local descent methods in federated learning","author":"Haddadpour","year":"2019"},{"key":"ref60","first-page":"13975","article-title":"Exponential graph is provably efficient for decentralized deep training","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"34","author":"Ying","year":"2021"},{"key":"ref61","first-page":"31717","article-title":"Stability and generalization of the decentralized stochastic gradient descent ascent algorithm","volume-title":"Proc. 37th Conf. Neural Inf. Process. Syst.","author":"Zhu","year":"2023"},{"key":"ref62","article-title":"Improved stability and generalization guarantees of the decentralized SGD algorithm","volume-title":"Proc. 41th Int. Conf. Mach. Learn.","author":"Bellet","year":"2024"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i14.29477"},{"key":"ref64","first-page":"31717","article-title":"Stability and generalization of the decentralized stochastic gradient descent ascent algorithm","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"36","author":"Zhu","year":"2024"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1137\/21M1465081"},{"key":"ref66","first-page":"2364","article-title":"Stability of SGD: Tightness analysis and improved bounds","volume-title":"Proc. Uncertainty Artif. Intell","author":"Zhang","year":"2022"},{"key":"ref67","article-title":"Lower generalization bounds for GD and SGD in smooth stochastic convex optimization","author":"Zhang","year":"2023"},{"key":"ref68","article-title":"Learning multiple layers of features from tiny images","author":"Krizhevsky","year":"2009"},{"issue":"7","key":"ref69","first-page":"3","article-title":"Tiny imagenet visual recognition challenge","volume":"7","author":"Le","year":"2015","journal-title":"CS 231 N"},{"key":"ref70","article-title":"Measuring the effects of non-identical data distribution for federated visual classification","author":"Hsu","year":"2019"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"}],"container-title":["IEEE Transactions on Pattern Analysis and Machine Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/34\/11424231\/11314751.pdf?arnumber=11314751","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,10]],"date-time":"2026-03-10T01:34:26Z","timestamp":1773106466000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11314751\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4]]},"references-count":71,"journal-issue":{"issue":"4"},"URL":"https:\/\/doi.org\/10.1109\/tpami.2025.3647762","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":[[2026,4]]}}}