{"id":1141,"date":"2021-07-18T21:54:35","date_gmt":"2021-07-18T12:54:35","guid":{"rendered":"https:\/\/wise.ajou.ac.kr:9605\/?page_id=1141"},"modified":"2021-11-26T16:11:53","modified_gmt":"2021-11-26T07:11:53","slug":"%eb%b0%b0%eb%af%bc%ed%98%b8","status":"publish","type":"page","link":"https:\/\/wise.ajou.ac.kr\/?page_id=1141","title":{"rendered":"\ubc30\ubbfc\ud638"},"content":{"rendered":"\n
\"\"<\/figure><\/div>\n\n\n\n

\ubc30\ubbfc\ud638
Minho Bae<\/sup><\/h1>\n\n\n\n

Email<\/h2>\n\n\n\n

minkkang1 at naver.com<\/p>\n\n\n\n

After graduation<\/h2>\n\n\n\n

\ud55c\uad6d\uc804\uc790\ud1b5\uc2e0\uc5f0\uad6c\uc6d0 (ETRI)<\/p>\n\n\n\n

Publications<\/h2>\n\n\n
<\/a><\/form>

2022<\/h3>
5.<\/div>

\uc5ec\uc0c1\ud638,; \ubc30\ubbfc\ud638,; \uc815\ubbfc\uc911,; \uad8c\uc624\uacbd,; \uc624\uc0c1\uc724,<\/p>

Crossover-SGD: A gossip-based communication in distributed deep learning for alleviating large mini-batch problem and enhancing scalability<\/a> Journal Article<\/span> <\/p>

In: <\/span>Concurrency and Computation: Practice and Experience, <\/span>2022<\/span>.<\/p>

Abstract<\/a><\/span> | Links<\/a><\/span> | BibTeX<\/a><\/span><\/p>

@article{\uc5ec\uc0c1\ud6382022Crossover-SGD,
\r\ntitle = {Crossover-SGD: A gossip-based communication in distributed deep learning for alleviating large mini-batch problem and enhancing scalability},
\r\nauthor = {\uc5ec\uc0c1\ud638 and \ubc30\ubbfc\ud638 and \uc815\ubbfc\uc911 and \uad8c\uc624\uacbd and \uc624\uc0c1\uc724},
\r\nurl = {https:\/\/arxiv.org\/abs\/2012.15198},
\r\ndoi = {10.48550\/arXiv.2012.15198},
\r\nyear  = {2022},
\r\ndate = {2022-11-01},
\r\nurldate = {2022-11-01},
\r\njournal = {Concurrency and Computation: Practice and Experience},
\r\nabstract = {     Distributed deep learning is an effective way to reduce the training time of deep learning for large datasets as well as complex models. However, the limited scalability caused by network overheads makes it difficult to synchronize the parameters of all workers. To resolve this problem, gossip-based methods that demonstrates stable scalability regardless of the number of workers have been proposed. However, to use gossip-based methods in general cases, the validation accuracy for a large mini-batch needs to be verified. To verify this, we first empirically study the characteristics of gossip methods in a large mini-batch problem and observe that the gossip methods preserve higher validation accuracy than AllReduce-SGD(Stochastic Gradient Descent) when the number of batch sizes is increased and the number of workers is fixed. However, the delayed parameter propagation of the gossip-based models decreases validation accuracy in large node scales. To cope with this problem, we propose Crossover-SGD that alleviates the delay propagation of weight parameters via segment-wise communication and load balancing random network topology. We also adapt hierarchical communication to limit the number of workers in gossip-based communication methods. To validate the effectiveness of our proposed method, we conduct empirical experiments and observe that our Crossover-SGD shows higher node scalability than SGP(Stochastic Gradient Push). },
\r\nkeywords = {},
\r\npubstate = {published},
\r\ntppubtype = {article}
\r\n}
\r\n<\/pre><\/div>
Close<\/a><\/p><\/div>
     Distributed deep learning is an effective way to reduce the training time of deep learning for large datasets as well as complex models. However, the limited scalability caused by network overheads makes it difficult to synchronize the parameters of all workers. To resolve this problem, gossip-based methods that demonstrates stable scalability regardless of the number of workers have been proposed. However, to use gossip-based methods in general cases, the validation accuracy for a large mini-batch needs to be verified. To verify this, we first empirically study the characteristics of gossip methods in a large mini-batch problem and observe that the gossip methods preserve higher validation accuracy than AllReduce-SGD(Stochastic Gradient Descent) when the number of batch sizes is increased and the number of workers is fixed. However, the delayed parameter propagation of the gossip-based models decreases validation accuracy in large node scales. To cope with this problem, we propose Crossover-SGD that alleviates the delay propagation of weight parameters via segment-wise communication and load balancing random network topology. We also adapt hierarchical communication to limit the number of workers in gossip-based communication methods. To validate the effectiveness of our proposed method, we conduct empirical experiments and observe that our Crossover-SGD shows higher node scalability than SGP(Stochastic Gradient Push). <\/div>
Close<\/a><\/p><\/div>