{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/elastic-gossip-distributing-neural-network","title":"Elastic Gossip: Distributing Neural Network Training Using Gossip-like Protocols","arxiv_id":"1812.02407","date":"2018-12-06","proceeding":null,"authors":["Siddharth Pramod"],"abstract":"Distributing Neural Network training is of particular interest for several\nreasons including scaling using computing clusters, training at data sources\nsuch as IOT devices and edge servers, utilizing underutilized resources across\nheterogeneous environments, and so on. Most contemporary approaches primarily\naddress scaling using computing clusters and require high network bandwidth and\nfrequent communication. This thesis presents an overview of standard approaches\nto distribute training and proposes a novel technique involving\npairwise-communication using Gossip-like protocols, called Elastic Gossip. This\napproach builds upon an existing technique known as Elastic Averaging SGD\n(EASGD), and is similar to another technique called Gossiping SGD which also\nuses Gossip-like protocols. Elastic Gossip is empirically evaluated against\nGossiping SGD using the MNIST digit recognition and CIFAR-10 classification\ntasks, using commonly used Neural Network architectures spanning Multi-Layer\nPerceptrons (MLPs) and Convolutional Neural Networks (CNNs). It is found that\nElastic Gossip, Gossiping SGD, and All-reduce SGD perform quite comparably,\neven though the latter entails a substantially higher communication cost. While\nElastic Gossip performs better than Gossiping SGD in these experiments, it is\npossible that a more thorough search over hyper-parameter space, specific to a\ngiven application, may yield configurations of Gossiping SGD that work better\nthan Elastic Gossip.","url_abs":"http://arxiv.org/abs/1812.02407v1","url_pdf":"http://arxiv.org/pdf/1812.02407v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"elastic-gossip-distributing-neural-network","repo_url":"https://github.com/sidps/dist_training","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[],"methods":[{"method_slug":"sgd","method_name":"SGD"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}