{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T01:20:55Z","timestamp":1781659255551,"version":"3.54.5"},"reference-count":101,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"11","license":[{"start":{"date-parts":[[2020,11,1]],"date-time":"2020-11-01T00:00:00Z","timestamp":1604188800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2020,11,1]],"date-time":"2020-11-01T00:00:00Z","timestamp":1604188800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2020,11,1]],"date-time":"2020-11-01T00:00:00Z","timestamp":1604188800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100004359","name":"Swedish Research Council","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100004359","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001729","name":"Swedish Foundation for Strategic Research","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100001729","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100004063","name":"Knut and Alice Wallenberg (KAW) Foundation","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100004063","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Proc. IEEE"],"published-print":{"date-parts":[[2020,11]]},"DOI":"10.1109\/jproc.2020.3026619","type":"journal-article","created":{"date-parts":[[2020,10,8]],"date-time":"2020-10-08T19:25:17Z","timestamp":1602185117000},"page":"2013-2031","source":"Crossref","is-referenced-by-count":62,"title":["Advances in Asynchronous Parallel and Distributed Optimization"],"prefix":"10.1109","volume":"108","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-9159-8447","authenticated-orcid":false,"given":"By Mahmoud","family":"Assran","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1725-2901","authenticated-orcid":false,"given":"Arda","family":"Aytekin","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hamid Reza","family":"Feyzmahdavian","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mikael","family":"Johansson","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0536-7904","authenticated-orcid":false,"given":"Michael G.","family":"Rabbat","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2016.2525015"},{"key":"ref38","first-page":"873","article-title":"Distributed delayed stochastic optimization","author":"agarwal","year":"2011","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1137\/080716542"},{"key":"ref32","first-page":"1017","article-title":"Stochastic gradient descent, weighted sampling, and the randomized Kaczmarz algorithm","author":"needell","year":"2014","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-540-24685-5_1"},{"key":"ref30","first-page":"2","article-title":"Parameter server for distributed machine learning","author":"li","year":"2013","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref37","first-page":"2331","article-title":"Slow learners are fast","author":"zinkevich","year":"2009","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1007\/s10107-012-0629-5"},{"key":"ref35","article-title":"Accelerated Bregman proximal gradient methods for relatively smooth convex optimization","author":"hanzely","year":"2018","journal-title":"arXiv 1808 03045"},{"key":"ref34","first-page":"819","article-title":"Non-uniform stochastic average gradient method for training conditional random fields","author":"schmidt","year":"2015","journal-title":"Proc Artif Intell Statist (AISTATS)"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/HiPC.2019.00032"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/IPDPSW.2016.116"},{"key":"ref29","first-page":"1223","article-title":"Large scale distributed deep networks","author":"dean","year":"2012","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref20","author":"goodfellow","year":"2016","journal-title":"Deep Learning"},{"key":"ref22","first-page":"165","article-title":"Optimal distributed online prediction using mini-batches","volume":"13","author":"dekel","year":"2012","journal-title":"J Mach Learn Res"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1214\/aoms\/1177729586"},{"key":"ref24","first-page":"590","article-title":"SySCD: A system-aware parallel coordinate descent algorithm","author":"ioannou","year":"2019","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/HPCC\/SmartCity\/DSS.2018.00116"},{"key":"ref101","doi-asserted-by":"publisher","DOI":"10.1109\/ICDM.2016.0074"},{"key":"ref26","article-title":"Improved parallel cache-oblivious algorithms for dynamic programming and linear algebra","author":"blleloch","year":"2018","journal-title":"arXiv 1809 09330"},{"key":"ref100","doi-asserted-by":"publisher","DOI":"10.1109\/CDC.2017.8264076"},{"key":"ref25","first-page":"1","article-title":"Measuring scheduling efficiency of RNNs for NLP applications","author":"thakker","year":"2019","journal-title":"Proc Int Workshop Perform Anal Mach Learn Syst (Fastpath)"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1137\/100802001"},{"key":"ref51","first-page":"321","article-title":"Parallel coordinate descent for L1-regularized loss minimization","author":"bradley","year":"2011","journal-title":"Proc Int Conf Mach Learn (ICML)"},{"key":"ref59","first-page":"2674","article-title":"Taming the wild: A unified analysis of Hogwild-style algorithms","author":"de sa","year":"2015","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref58","first-page":"2737","article-title":"Asynchronous parallel stochastic gradient for nonconvex optimization","author":"lian","year":"2015","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref57","first-page":"56","article-title":"Breaking the nonsmooth barrier: A scalable parallel method for composite optimization","author":"pedregosa","year":"2017","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref56","article-title":"Variance reduced coordinate descent with acceleration: New method with a surprising application to finite-sum problems","author":"hanzely","year":"2020","journal-title":"arXiv 2002 04670"},{"key":"ref55","first-page":"2082","article-title":"SEGA: Variance reduction via gradient sketching","author":"hanzely","year":"2018","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1137\/15M1024950"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1137\/140961134"},{"key":"ref52","doi-asserted-by":"crossref","first-page":"433","DOI":"10.1007\/s10107-015-0901-6","article-title":"Parallel coordinate descent methods for big data optimization","volume":"156","author":"richt\u00e1rik","year":"2015","journal-title":"Math Program"},{"key":"ref40","first-page":"693","article-title":"Hogwild: A lock-free approach to parallelizing stochastic gradient descent","author":"recht","year":"2011","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1016\/0024-3795(69)90028-7"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01216-8_12"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1145\/322063.322067"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/363626.363628"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1007\/BF02591967"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1007\/BF00939948"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1137\/0325006"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.1986.1104412"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1137\/040615961"},{"key":"ref45","first-page":"378","article-title":"Accelerated mini-batch stochastic dual coordinate ascent","author":"shalev-shwartz","year":"2013","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1137\/16M1057000"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1137\/15M1049695"},{"key":"ref42","first-page":"2663","article-title":"A stochastic gradient method with an exponential convergence rate for finite training sets","author":"le roux","year":"2012","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref41","first-page":"46","article-title":"ASAGA: Asynchronous parallel SAGA","volume":"54","author":"leblond","year":"2017","journal-title":"Proc Int Conf Artif Intell Statist (AISTATS)"},{"key":"ref44","first-page":"1646","article-title":"SAGA: A fast incremental gradient method with support for non-strongly convex composite objectives","author":"defazio","year":"2014","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref43","first-page":"315","article-title":"Accelerating stochastic gradient descent using predictive variance reduction","author":"johnson","year":"2013","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref73","first-page":"344","article-title":"Stochastic gradient push for distributed deep learning","volume":"97","author":"assran","year":"2019","journal-title":"Proc Int Conf Mach Learn (ICML)"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2016.2529285"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1109\/SFCS.2003.1238221"},{"key":"ref70","first-page":"1","article-title":"Revisiting distributed synchronous SGD","author":"chen","year":"2016","journal-title":"Proc Int Conf Learn Represent (ICLR)"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2019.2930234"},{"key":"ref77","article-title":"Robust asynchronous stochastic gradient-push: Asymptotically optimal and network-independent performance for strongly convex functions","author":"spiridonoff","year":"2018","journal-title":"arXiv 1811 03982"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1109\/GlobalSIP.2017.8309024"},{"key":"ref75","article-title":"Asynchronous gradient-push","author":"assran","year":"2018","journal-title":"arXiv 1803 08950"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1109\/ALLERTON.2018.8636055"},{"key":"ref79","article-title":"Asynchronous decentralized successive convex approximation","author":"tian","year":"2019","journal-title":"arXiv 1909 10144"},{"key":"ref60","first-page":"3750","article-title":"SGD and Hogwild! convergence without the bounded gradients assumption","volume":"80","author":"nguyen","year":"2018","journal-title":"Proc Int Conf Mach Learn (ICML)"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1109\/MLSP.2014.6958872"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1007\/s10957-013-0409-2"},{"key":"ref63","article-title":"Analysis and implementation of an asynchronous optimization algorithm for the parameter server","author":"aytekin","year":"2016","journal-title":"arXiv 1610 05507"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1109\/CDC.2014.7039374"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1137\/140957639"},{"key":"ref66","first-page":"1125","article-title":"Finito: A faster, permutable incremental gradient method for big data problems","author":"defazio","year":"2014","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1137\/16M1101702"},{"key":"ref68","first-page":"3587","article-title":"A delay-tolerant proximal-gradient algorithm for distributed learning","author":"mishchenko","year":"2018","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref2","article-title":"RoBERTa: A robustly optimized BERT pretraining approach","author":"liu","year":"2019","journal-title":"arXiv 1907 11692"},{"key":"ref69","first-page":"1111","article-title":"A reliable effective terascale linear learning system","volume":"15","author":"agarwal","year":"2014","journal-title":"J Mach Learn Res"},{"key":"ref1","article-title":"Language models are unsupervised multi-task learners","author":"radford","year":"2019"},{"key":"ref95","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2018.2817461"},{"key":"ref94","article-title":"SlowMo: Improving communication-efficient distributed SGD with slow momentum","author":"wang","year":"2019","journal-title":"arXiv 1910 00643"},{"key":"ref93","first-page":"3043","article-title":"Asynchronous decentralized parallel stochastic gradient descent","volume":"80","author":"lian","year":"2018","journal-title":"Proc Int Conf Mach Learn (ICML)"},{"key":"ref92","first-page":"5330","article-title":"Can decentralized algorithms outperform centralized algorithms? A case study for decentralized parallel stochastic gradient descent","author":"lian","year":"2017","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref91","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-015-0816-y"},{"key":"ref90","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref98","article-title":"Lecture notes in neural networks for machine learning","author":"hinton","year":"2014"},{"key":"ref99","doi-asserted-by":"publisher","DOI":"10.1007\/s10107-019-01408-w"},{"key":"ref96","article-title":"Asymptotic network independence in distributed stochastic optimization for machine learning","author":"pu","year":"2019","journal-title":"arXiv 1906 12345"},{"key":"ref97","article-title":"Adam: A method for stochastic optimization","author":"kingma","year":"2014","journal-title":"arXiv 1412 6980"},{"key":"ref10","volume":"23","author":"bertsekas","year":"1989","journal-title":"Parallel and Distributed Computation"},{"key":"ref11","first-page":"161","article-title":"The tradeoffs of large scale learning","author":"bottou","year":"2007","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1137\/16M1080173"},{"key":"ref13","article-title":"Stochastic Newton and cubic Newton methods with simple local linear-quadratic rates","author":"kovalev","year":"2019","journal-title":"arXiv 1912 01597"},{"key":"ref14","first-page":"1965","article-title":"DAve-QN: A distributed averaged quasi-Newton method with local superlinear convergence rate","author":"soori","year":"2020","journal-title":"Proc Int Conf Artif Intell Statist"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/TPWRS.2019.2909664"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.2008.924723"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1137\/060657005"},{"key":"ref17","doi-asserted-by":"crossref","DOI":"10.1007\/978-1-4419-8853-9","author":"nesterov","year":"2004","journal-title":"Introductory Lectures on Convex Optimization"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.1109\/ECC.2015.7330626"},{"key":"ref18","author":"murphy","year":"2012","journal-title":"Machine Learning A Probabilistic Perspective"},{"key":"ref84","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2013.2275669"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1017\/CBO9781107298019"},{"key":"ref83","doi-asserted-by":"publisher","DOI":"10.1109\/Allerton.2011.6120272"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.1109\/TAC.2015.2512043"},{"key":"ref89","first-page":"1","article-title":"Automatic differentiation in PyTorch","author":"paszke","year":"2017","journal-title":"Proc Workshop Automatic Differentiation"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.1109\/TCNS.2015.2426732"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.2307\/2034984"},{"key":"ref87","doi-asserted-by":"publisher","DOI":"10.1007\/0-387-32792-4"},{"key":"ref88","doi-asserted-by":"publisher","DOI":"10.1137\/16M1084316"}],"container-title":["Proceedings of the IEEE"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/5\/9241485\/09217472.pdf?arnumber=9217472","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,4,27]],"date-time":"2022-04-27T13:39:53Z","timestamp":1651066793000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9217472\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,11]]},"references-count":101,"journal-issue":{"issue":"11"},"URL":"https:\/\/doi.org\/10.1109\/jproc.2020.3026619","relation":{},"ISSN":["0018-9219","1558-2256"],"issn-type":[{"value":"0018-9219","type":"print"},{"value":"1558-2256","type":"electronic"}],"subject":[],"published":{"date-parts":[[2020,11]]}}}