{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,9]],"date-time":"2026-01-09T20:32:23Z","timestamp":1767990743280,"version":"3.49.0"},"reference-count":37,"publisher":"IEEE","license":[{"start":{"date-parts":[[2021,5,10]],"date-time":"2021-05-10T00:00:00Z","timestamp":1620604800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2021,5,10]],"date-time":"2021-05-10T00:00:00Z","timestamp":1620604800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2021,5,10]],"date-time":"2021-05-10T00:00:00Z","timestamp":1620604800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2021,5,10]]},"DOI":"10.1109\/infocom42981.2021.9488678","type":"proceedings-article","created":{"date-parts":[[2021,7,27]],"date-time":"2021-07-27T00:07:32Z","timestamp":1627344452000},"page":"1-10","source":"Crossref","is-referenced-by-count":13,"title":["Near-Optimal Topology-adaptive Parameter Synchronization in Distributed DNN Training"],"prefix":"10.1109","author":[{"given":"Zhe","family":"Zhang","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Chuan","family":"Wu","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zongpeng","family":"Li","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref33","article-title":"Asynchronous decentralized parallel stochastic gradient descent","author":"lian","year":"2018","journal-title":"Proceedings of the 35th International Conference on Machine Learning ICML 2018 Stockholmsmässan"},{"key":"ref32","article-title":"Plink: Discovering and exploiting locality for accelerated distributed training on the public cloud","author":"luo","year":"2020","journal-title":"Proceedings of MLSys"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/ICC47138.2019.9123209"},{"key":"ref30","article-title":"BML: A high-performance, low-cost gradient synchronization algorithm for DML training","author":"wang","year":"2018","journal-title":"NeurIPS"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1007\/s10878-016-0074-0"},{"key":"ref36","article-title":"Packing steiner trees","author":"jain","year":"2003","journal-title":"Proceedings of the fourteenth annual ACM-SIAM symposium on Discrete algorithms"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1007\/BF01585866"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1007\/s00453-005-1188-4"},{"key":"ref10","article-title":"Highly scalable deep learning training system with mixed-precision: Training imagenet in four minutes","author":"jia","year":"2018","journal-title":"CoRR"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1147\/JRD.2019.2947013"},{"key":"ref12","article-title":"Image classification at supercomputer scale","author":"ying","year":"2018","journal-title":"CoRR"},{"key":"ref13","article-title":"Imagenet\/resnet-50 training in 224 seconds","author":"mikami","year":"2018","journal-title":"CoRR"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1145\/1402958.1402967"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1145\/1592568.1592577"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359642"},{"key":"ref17","article-title":"Horovod: fast and easy distributed deep learning in tensorflow","author":"sergeev","year":"2018","journal-title":"CoRR"},{"key":"ref18","doi-asserted-by":"crossref","DOI":"10.1016\/j.jpdc.2008.09.002","article-title":"Bandwidth optimal all-reduce algorithms for clusters of workstations","author":"patarasuk","year":"2009","journal-title":"J Parallel Distributed Comput"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1016\/j.parco.2009.09.001"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1007\/s11036-013-0489-0"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.14778\/2733004.2733082"},{"key":"ref27","article-title":"DGX-2","year":"0"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1177\/0278364917710318"},{"key":"ref6","article-title":"Scaling distributed machine learning with the parameter server","author":"li","year":"2014","journal-title":"USENIX Symp on Operating Systems Design & Implementation"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1145\/1402958.1402968"},{"key":"ref5","article-title":"Bringing hpc techniques to deep learning","author":"gibiansky","year":"2017"},{"key":"ref8","article-title":"Blink: Fast and generic collectives for distributed ML","author":"wang","year":"2020","journal-title":"Proceedings of MLSys"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359646"},{"key":"ref2","article-title":"BERT: pre-training of deep bidirectional transformers for language understanding","author":"devlin","year":"2019","journal-title":"NAACL-HLT"},{"key":"ref9","doi-asserted-by":"crossref","first-page":"401","DOI":"10.1145\/3373376.3378499","author":"luo","year":"2020","journal-title":"Proc ASPLOS Architectural Support for Programming Languages and Operating Systems"},{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref20","article-title":"NCCL 2.4","year":"0"},{"key":"ref22","article-title":"NVLink","year":"0"},{"key":"ref21","article-title":"PCIe 5.0","year":"0"},{"key":"ref24","article-title":"GPUDirect RDMA","year":"0"},{"key":"ref23","article-title":"InfiniBand","year":"0"},{"key":"ref26","article-title":"DGX-1","year":"0"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2019.2928289"}],"event":{"name":"IEEE INFOCOM 2021 - IEEE Conference on Computer Communications","location":"Vancouver, BC, Canada","start":{"date-parts":[[2021,5,10]]},"end":{"date-parts":[[2021,5,13]]}},"container-title":["IEEE INFOCOM 2021 - IEEE Conference on Computer Communications"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/9488422\/9488423\/09488678.pdf?arnumber=9488678","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,5,10]],"date-time":"2022-05-10T15:43:36Z","timestamp":1652197416000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9488678\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,5,10]]},"references-count":37,"URL":"https:\/\/doi.org\/10.1109\/infocom42981.2021.9488678","relation":{},"subject":[],"published":{"date-parts":[[2021,5,10]]}}}