{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,18]],"date-time":"2026-07-18T16:03:09Z","timestamp":1784390589634,"version":"3.55.0"},"reference-count":81,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"1","license":[{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"am","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"Wei Shen and Xuehong Zhang Presidential Fellowship at Carnegie Mellon University"},{"DOI":"10.13039\/100000006","name":"ONR","doi-asserted-by":"publisher","award":["N00014-19-1-2404"],"award-info":[{"award-number":["N00014-19-1-2404"]}],"id":[{"id":"10.13039\/100000006","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100006602","name":"Air Force Research Laboratory","doi-asserted-by":"publisher","award":["FA8750-20-2-0504"],"award-info":[{"award-number":["FA8750-20-2-0504"]}],"id":[{"id":"10.13039\/100006602","id-type":"DOI","asserted-by":"publisher"}]},{"name":"NSF","award":["CCF-1901199"],"award-info":[{"award-number":["CCF-1901199"]}]},{"name":"NSF","award":["CCF-2007911"],"award-info":[{"award-number":["CCF-2007911"]}]},{"name":"NSF","award":["CNS-2148212"],"award-info":[{"award-number":["CNS-2148212"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE J. Sel. Top. Signal Process."],"published-print":{"date-parts":[[2025,1]]},"DOI":"10.1109\/jstsp.2025.3526081","type":"journal-article","created":{"date-parts":[[2025,1,3]],"date-time":"2025-01-03T19:25:42Z","timestamp":1735932342000},"page":"273-282","source":"Crossref","is-referenced-by-count":5,"title":["Convergence and Privacy of Decentralized Nonconvex Optimization With Gradient Clipping and Communication Compression"],"prefix":"10.1109","volume":"19","author":[{"given":"Boyue","family":"Li","sequence":"first","affiliation":[{"name":"Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6766-5459","authenticated-orcid":false,"given":"Yuejie","family":"Chi","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","first-page":"3478","article-title":"Decentralized stochastic optimization and gossip algorithms with compressed communication","volume-title":"Proc. 36th Int. Conf. Mach. Learn.","author":"Koloskova","year":"2019"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/JSAIT.2021.3103920"},{"key":"ref3","article-title":"BEER: Fast O (1\/T) rate for decentralized nonconvex optimization with communication compression","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhao","year":"2022"},{"key":"ref4","article-title":"Deepsqueeze: Decentralization meets error-compensated compression","author":"Tang","year":"2019"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2024.3471854"},{"key":"ref6","first-page":"39035","article-title":"Compressed decentralized proximal stochastic gradient method for nonconvex composite problems with heterogeneous data","volume-title":"Proc. 40th Int. Conf. Mach. Learn.","author":"Yan","year":"2023"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1145\/2976749.2978318"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-540-79228-4_1"},{"key":"ref9","article-title":"Why gradient clipping accelerates training: A theoretical justification for adaptivity","author":"Zhang","year":"2019"},{"key":"ref10","first-page":"2203","article-title":"Efficient privacy-preserving stochastic nonconvex optimization","volume-title":"Proc. Conf. Uncertainty Artif. Intell.","author":"Wang","year":"2023"},{"key":"ref11","first-page":"4285","article-title":"SoteriaFL: A unified framework for private federated learning with communication compression","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Li","year":"2022"},{"key":"ref12","article-title":"Normalized\/clipped SGD with perturbation for differentially private non-convex optimization","author":"Yang","year":"2022"},{"key":"ref13","first-page":"15511","article-title":"Improved analysis of clipping algorithms for non-convex optimization","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhang","year":"2020"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2020.3021381"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2020.3024266"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/SFCS.2003.1238221"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1016\/j.sysconle.2004.02.022"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1561\/1300000014"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2012.2209984"},{"key":"ref20","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.","author":"Lian","year":"2017"},{"key":"ref21","first-page":"344","article-title":"Stochastic gradient push for distributed deep learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Assran","year":"2019"},{"issue":"1","key":"ref22","first-page":"213:9709","article-title":"Cooperative SGD: A unified framework for the design and analysis of local-update SGD algorithms","volume":"22","author":"Wang","year":"2021","journal-title":"J. Mach. Learn. Res."},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1016\/j.automatica.2009.10.021"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/TSIPN.2016.2524588"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1137\/16M1084316"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/TCNS.2017.2698261"},{"issue":"180","key":"ref27","first-page":"1","article-title":"Communication-efficient distributed optimization in networks with gradient tracking and variance reduction","volume":"21","author":"Li","year":"2020","journal-title":"J. Mach. Learn. Res."},{"key":"ref28","first-page":"9217","article-title":"Improving the sample and communication complexity for decentralized non-convex optimization: Joint gradient estimation and tracking","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Sun","year":"2020"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2021.3122586"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1137\/20M1361158"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1137\/21M1450677"},{"key":"ref32","first-page":"9310","article-title":"Tackling data heterogeneity: A new unified framework for decentralized SGD with sample-induced topology","volume-title":"Proc. 39th Int. Conf. Mach. Learn.","author":"Huang","year":"2022"},{"key":"ref33","article-title":"An optimal stochastic algorithm for decentralized nonconvex finite-sum optimization","author":"Luo","year":"2022"},{"key":"ref34","first-page":"2674","article-title":"Taming the wild: A unified analysis of HOG WILD-style algorithms","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Sa","year":"2015"},{"key":"ref35","first-page":"1707","article-title":"QSGD: Communication-efficient SGD via gradient quantization and encoding","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Alistarh","year":"2017"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2014-274"},{"key":"ref37","first-page":"4452","article-title":"Sparsified SGD with memory","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Stich","year":"2018"},{"key":"ref38","first-page":"5977","article-title":"The convergence of sparsified gradient methods","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Alistarh","year":"2018"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1080\/10556788.2024.2358790"},{"key":"ref40","first-page":"5895","article-title":"Acceleration for compressed gradient descent in distributed and federated optimization","volume-title":"Proc. 37th Int. Conf. Mach. Learn.","author":"Li","year":"2020"},{"key":"ref41","first-page":"3788","article-title":"MARINA: Faster non-convex distributed learning with compression","volume-title":"Proc. 38th Int. Conf. Mach. Learn.","author":"Gorbunov","year":"2021"},{"key":"ref42","first-page":"13770","article-title":"CANITA: Faster rates for distributed convex optimization with communication compression","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Li","year":"2021"},{"key":"ref43","first-page":"4384","article-title":"EF21: A new, simpler, theoretically better, and practically faster error feedback","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Richtarik","year":"2021"},{"key":"ref44","article-title":"EF21 with bells & whistles: Practical algorithmic extensions of modern error feedback","author":"Fatkhullin","year":"2021"},{"key":"ref45","first-page":"7652","article-title":"Communication compression for decentralized training","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Tang","year":"2018"},{"key":"ref46","first-page":"9324","article-title":"Quantized decentralized stochastic learning over directed graphs","volume-title":"Proc. 37th Int. Conf. Mach. Learn.","author":"Taheri","year":"2020"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/CDC49753.2023.10384180"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1137\/23M1553820"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.23919\/EUSIPCO63174.2024.10715377"},{"key":"ref50","first-page":"2701","article-title":"Escaping saddle points in heterogeneous federated learning via distributed SGD with communication compression","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Chen","year":"2024"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1007\/11681878_14"},{"key":"ref52","first-page":"2719","article-title":"Differentially private empirical risk minimization revisited: Faster and more general","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Wang","year":"2017"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/SP.2019.00001"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1145\/3357713.3384335"},{"key":"ref55","first-page":"13773","article-title":"Understanding gradient clipping in private SGD: A geometric perspective","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Chen","year":"2020"},{"key":"ref56","first-page":"10081","article-title":"On differentially private stochastic convex optimization with heavy-tailed data","volume-title":"Proc. 37th Int. Conf. Mach. Learn.","author":"Wang","year":"2020"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.1007\/s11280-020-00780-4"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/ISIT45174.2021.9518124"},{"key":"ref59","first-page":"10110","article-title":"Differentially private federated learning on heterogeneous data","volume-title":"Proc. 25th Int. Conf. Artif. Intell. Statist.","author":"Noble","year":"2022"},{"key":"ref60","article-title":"Dynamic differential-privacy preserving SGD","author":"Du","year":"2022"},{"key":"ref61","first-page":"25523","article-title":"DIFF2: Differential private optimization via gradient differences for nonconvex distributed learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Murata","year":"2023"},{"key":"ref62","first-page":"26048","article-title":"Understanding clipping for federated learning: Convergence and client-level differential privacy","volume-title":"Proc. 39th Int. Conf. Mach. Learn.","author":"Zhang","year":"2022"},{"key":"ref63","first-page":"1310","article-title":"On the difficulty of training recurrent neural networks","volume-title":"Proc. 30th Int. Conf. Mach. Learn.","author":"Pascanu","year":"2013"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2013.6639349"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.182"},{"key":"ref66","first-page":"435","article-title":"Preventing gradient explosions in gated recurrent units","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Kanai","year":"2017"},{"issue":"UCB\/EECS-2017-156","key":"ref67","article-title":"Scaling SGD batch size to 32k for ImageNet training","author":"You","year":"2017"},{"key":"ref68","article-title":"Why gradient clipping accelerates training: A theoretical justification for adaptivity","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Zhang","year":"2020"},{"key":"ref69","first-page":"15383","article-title":"Why are adaptive methods good for attention models?","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhang","year":"2020"},{"key":"ref70","article-title":"Variance-reduced clipping for non-convex optimization","author":"Reisizadeh","year":"2023"},{"key":"ref71","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":"ref72","article-title":"Improved convergence of differential private SGD with gradient clipping","volume-title":"Proc. 11th Int. Conf. Learn. Representations","author":"Fang","year":"2023"},{"key":"ref73","article-title":"Revisiting gradient clipping: Stochastic bias and tight convergence guarantees","volume-title":"Proc. 40th Int. Conf. Mach. Learn.","author":"Koloskova","year":"2023"},{"key":"ref74","article-title":"Convergence and privacy of decentralized nonconvex optimization with gradient clipping and communication compression","author":"Li","year":"2023"},{"key":"ref75","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2018.2817461"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1109\/FOCS.2013.53"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-39077-7_5"},{"key":"ref78","article-title":"Local differential privacy in decentralized optimization","author":"Xiao","year":"2019"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.1145\/1961189.1961199"},{"key":"ref80","first-page":"2406","article-title":"SpiderBoost and momentum: Faster variance reduction algorithms","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Wang","year":"2019"},{"issue":"276","key":"ref81","first-page":"2","article-title":"Learning algorithms for classification: A comparison on handwritten digit recognition","volume":"261","author":"LeCun","year":"1995","journal-title":"Neural Netw., Statist. Mechan. Perspective"}],"container-title":["IEEE Journal of Selected Topics in Signal Processing"],"original-title":[],"link":[{"URL":"https:\/\/ieeexplore.ieee.org\/ielam\/4200690\/10906680\/10824895-aam.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/4200690\/10906680\/10824895.pdf?arnumber=10824895","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,12]],"date-time":"2025-06-12T17:42:49Z","timestamp":1749750169000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10824895\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,1]]},"references-count":81,"journal-issue":{"issue":"1"},"URL":"https:\/\/doi.org\/10.1109\/jstsp.2025.3526081","relation":{},"ISSN":["1932-4553","1941-0484"],"issn-type":[{"value":"1932-4553","type":"print"},{"value":"1941-0484","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,1]]}}}