{"publisher":"Journal of Machine Learning Research","scopus_import":"1","oa":1,"oa_version":"Published Version","date_created":"2021-06-20T22:01:33Z","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","quality_controlled":"1","title":"NUQSGD: Provably communication-efficient data-parallel SGD via nonuniform quantization","status":"public","month":"04","article_processing_charge":"No","tmp":{"image":"/images/cc_by.png","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"date_updated":"2024-03-06T12:22:07Z","file":[{"file_size":11237154,"date_created":"2021-06-23T07:09:41Z","success":1,"file_id":"9595","date_updated":"2021-06-23T07:09:41Z","creator":"asandaue","access_level":"open_access","file_name":"2021_JournalOfMachineLearningResearch_Ramezani-Kebrya.pdf","content_type":"application/pdf","checksum":"6428aa8bcb67768b6949c99b55d5281d","relation":"main_file"}],"type":"journal_article","day":"01","department":[{"_id":"DaAl"}],"abstract":[{"text":"As the size and complexity of models and datasets grow, so does the need for communication-efficient variants of stochastic gradient descent that can be deployed to perform parallel model training. One popular communication-compression method for data-parallel SGD is QSGD (Alistarh et al., 2017), which quantizes and encodes gradients to reduce communication costs. The baseline variant of QSGD provides strong theoretical guarantees, however, for practical purposes, the authors proposed a heuristic variant which we call QSGDinf, which demonstrated impressive empirical gains for distributed training of large neural networks. In this paper, we build on this work to propose a new gradient quantization scheme, and show that it has both stronger theoretical guarantees than QSGD, and matches and exceeds the empirical performance of the QSGDinf heuristic and of other compression methods.","lang":"eng"}],"main_file_link":[{"open_access":"1","url":"https://www.jmlr.org/papers/v22/20-255.html"}],"date_published":"2021-04-01T00:00:00Z","publication_status":"published","issue":"114","year":"2021","intvolume":"        22","volume":22,"ddc":["000"],"file_date_updated":"2021-06-23T07:09:41Z","external_id":{"arxiv":["1908.06077"]},"publication_identifier":{"issn":["15324435"],"eissn":["15337928"]},"page":"1−43","has_accepted_license":"1","author":[{"full_name":"Ramezani-Kebrya, Ali","last_name":"Ramezani-Kebrya","first_name":"Ali"},{"full_name":"Faghri, Fartash","last_name":"Faghri","first_name":"Fartash"},{"full_name":"Markov, Ilya","last_name":"Markov","first_name":"Ilya"},{"full_name":"Aksenov, Vitalii","id":"2980135A-F248-11E8-B48F-1D18A9856A87","last_name":"Aksenov","first_name":"Vitalii"},{"id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","full_name":"Alistarh, Dan-Adrian","last_name":"Alistarh","first_name":"Dan-Adrian","orcid":"0000-0003-3650-940X"},{"full_name":"Roy, Daniel M.","first_name":"Daniel M.","last_name":"Roy"}],"publication":"Journal of Machine Learning Research","language":[{"iso":"eng"}],"_id":"9571","article_type":"original","citation":{"short":"A. Ramezani-Kebrya, F. Faghri, I. Markov, V. Aksenov, D.-A. Alistarh, D.M. Roy, Journal of Machine Learning Research 22 (2021) 1−43.","apa":"Ramezani-Kebrya, A., Faghri, F., Markov, I., Aksenov, V., Alistarh, D.-A., &#38; Roy, D. M. (2021). NUQSGD: Provably communication-efficient data-parallel SGD via nonuniform quantization. <i>Journal of Machine Learning Research</i>. Journal of Machine Learning Research.","ista":"Ramezani-Kebrya A, Faghri F, Markov I, Aksenov V, Alistarh D-A, Roy DM. 2021. NUQSGD: Provably communication-efficient data-parallel SGD via nonuniform quantization. Journal of Machine Learning Research. 22(114), 1−43.","chicago":"Ramezani-Kebrya, Ali, Fartash Faghri, Ilya Markov, Vitalii Aksenov, Dan-Adrian Alistarh, and Daniel M. Roy. “NUQSGD: Provably Communication-Efficient Data-Parallel SGD via Nonuniform Quantization.” <i>Journal of Machine Learning Research</i>. Journal of Machine Learning Research, 2021.","ama":"Ramezani-Kebrya A, Faghri F, Markov I, Aksenov V, Alistarh D-A, Roy DM. NUQSGD: Provably communication-efficient data-parallel SGD via nonuniform quantization. <i>Journal of Machine Learning Research</i>. 2021;22(114):1−43.","ieee":"A. Ramezani-Kebrya, F. Faghri, I. Markov, V. Aksenov, D.-A. Alistarh, and D. M. Roy, “NUQSGD: Provably communication-efficient data-parallel SGD via nonuniform quantization,” <i>Journal of Machine Learning Research</i>, vol. 22, no. 114. Journal of Machine Learning Research, p. 1−43, 2021.","mla":"Ramezani-Kebrya, Ali, et al. “NUQSGD: Provably Communication-Efficient Data-Parallel SGD via Nonuniform Quantization.” <i>Journal of Machine Learning Research</i>, vol. 22, no. 114, Journal of Machine Learning Research, 2021, p. 1−43."}}