{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T15:44:22Z","timestamp":1782834262139,"version":"3.54.5"},"reference-count":47,"publisher":"IBM","issue":"6","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IBM J. Res. & Dev."],"published-print":{"date-parts":[[2019,11,1]]},"DOI":"10.1147\/jrd.2019.2947013","type":"journal-article","created":{"date-parts":[[2019,10,14]],"date-time":"2019-10-14T15:26:08Z","timestamp":1571066768000},"page":"1:1-1:11","source":"Crossref","is-referenced-by-count":68,"title":["BlueConnect: Decomposing all-reduce for deep learning on heterogeneous network hierarchy"],"prefix":"10.1147","volume":"63","author":[{"given":"M.","family":"Cho","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"U.","family":"Finkler","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"M.","family":"Serrano","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"D.","family":"Kung","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"H.","family":"Hunter","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"3082","reference":[{"key":"ref39","year":"0"},{"key":"ref38","first-page":"95","author":"dichev","year":"2014","journal-title":"Optimization of Collective Communication for Heterogeneous HPC Platforms"},{"key":"ref33","article-title":"Cooperative SGD: A unified framework for the design and analysis of communication-efficient SGD algorithms","author":"wang","year":"2018","journal-title":"CoRR"},{"key":"ref32","article-title":"Distributed training of deep neuronal networks: Theoreticaland practical limits of parallel scalability","author":"keuper","year":"2016","journal-title":"CoRR"},{"key":"ref31","year":"2017","journal-title":"Wikipedia"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1177\/1094342005051521"},{"key":"ref37","first-page":"5336","article-title":"Can decentralized algorithms outperform centralized algorithms? a case study for decentralized parallel stochastic gradient descent","volume":"30","author":"lian","year":"2017","journal-title":"Adv Neural Inf Process Syst"},{"key":"ref36","first-page":"2350","article-title":"Staleness-aware async- SGD for distributed deep learning","author":"zhang","year":"2016","journal-title":"Proc Int Joint Conf Artif Intell"},{"key":"ref35","first-page":"685","article-title":"Deep learning with elastic averaging SGD","volume":"1","author":"zhang","year":"0","journal-title":"Proc Int Conf Neural Inf Process"},{"key":"ref34","first-page":"693","article-title":"Hogwild!: A lock-free approach to parallelizing stochastic gradient descent","author":"niu","year":"2011","journal-title":"Proc Int Conf Neural Inf Process"},{"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":"ref40","doi-asserted-by":"publisher","DOI":"10.1145\/1402946.1402967"},{"key":"ref11","article-title":"Scaling SGD batch size to 32k for ImageNet training","author":"you","year":"2017","journal-title":"CoRR"},{"key":"ref12","first-page":"1:1","article-title":"ImageNet training in minutes","author":"you","year":"2018","journal-title":"Proc Int Conf Parallel Process"},{"key":"ref13","year":"2017"},{"key":"ref14","article-title":"Coupling adaptive batch sizes with learning rates","author":"balles","year":"2017","journal-title":"Proc Conf Uncertainty Artif Intell"},{"key":"ref15","article-title":"On large-batch training for deep learning: Generalization gap and sharp minima","author":"keskar","year":"2017","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref16","article-title":"One weird trick for parallelizing convolutional neural networks","author":"krizhevsky","year":"2014","journal-title":"CoRR"},{"key":"ref17","article-title":"On scale-out deep learning training for cloud and HPC","author":"sridharan","year":"2018","journal-title":"CoRR"},{"key":"ref18","first-page":"265","article-title":"Tensorflow: A system for large-scale machine learning","author":"abadi","year":"2016","journal-title":"Proc USENIX Symp Operat Syst Des Implement"},{"key":"ref19","article-title":"Mxnet: A flexible and efficient machine learning library for heterogeneous distributed systems","author":"chen","year":"2015","journal-title":"CoRR"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1145\/1088149.1088183"},{"key":"ref4","article-title":"Revisiting distributed synchronous SGD","author":"chen","year":"2016","journal-title":"Proc Workshop Int Conf Learn Represent"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1145\/2647868.2654889"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1145\/3079856.3080246"},{"key":"ref29","year":"2017"},{"key":"ref5","first-page":"1097","article-title":"ImageNet classification with deep convolutional neural networks","volume":"25","author":"krizhevsky","year":"2012","journal-title":"Adv Neural Inf Process Syst"},{"key":"ref8","article-title":"Accurate, large mini batch SGD: training ImageNet in 1 hour","author":"goyal","year":"2017","journal-title":"CoRR"},{"key":"ref7","year":"2017"},{"key":"ref2","first-page":"448","article-title":"Batch normalization: Accelerating deep network training by reducing internal covariate shift","volume":"37","author":"ioffe","year":"0","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref9","article-title":"Firecaffe: Near-linear acceleration of deep neural network training on compute clusters","author":"iandola","year":"2015","journal-title":"Proc IEEE Comput Vis Pattern Recogn"},{"key":"ref1","first-page":"173","article-title":"Deep speech 2: End-to-end speech recognition in english and mandarin","author":"amodei","year":"2016","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref46","article-title":"Automated design of torus networks","author":"solnushkin","year":"2013","journal-title":"CoRR"},{"key":"ref20","year":"0"},{"key":"ref45","year":"0"},{"key":"ref22","article-title":"ChainerC: A library for deep learning in computer vision","author":"niitani","year":"2017","journal-title":"Porc Int Conf Multimedia"},{"key":"ref47","year":"2017"},{"key":"ref21","year":"0"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/SC.2000.10040"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1145\/2939672.2945397"},{"key":"ref41","year":"0"},{"key":"ref23","year":"2017"},{"key":"ref44","year":"0"},{"key":"ref26","year":"2017"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/TCC.2015.2415804"},{"key":"ref25","year":"2017"}],"container-title":["IBM Journal of Research and Development"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/5288520\/8894910\/08868152.pdf?arnumber=8868152","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,20]],"date-time":"2025-10-20T17:55:28Z","timestamp":1760982928000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/8868152\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2019,11,1]]},"references-count":47,"journal-issue":{"issue":"6"},"URL":"https:\/\/doi.org\/10.1147\/jrd.2019.2947013","relation":{},"ISSN":["0018-8646","0018-8646"],"issn-type":[{"value":"0018-8646","type":"print"},{"value":"0018-8646","type":"electronic"}],"subject":[],"published":{"date-parts":[[2019,11,1]]}}}