{"id":851,"date":"2021-06-30T11:47:41","date_gmt":"2021-06-30T02:47:41","guid":{"rendered":"https:\/\/wise.ajou.ac.kr:9605\/?page_id=851"},"modified":"2023-08-15T11:05:34","modified_gmt":"2023-08-15T02:05:34","slug":"%ec%9c%a4%eb%8c%80%ea%b1%b4","status":"publish","type":"page","link":"https:\/\/wise.ajou.ac.kr\/?page_id=851","title":{"rendered":"\uc724\ub300\uac74"},"content":{"rendered":"<div class=\"wp-block-image is-style-default\">\n<figure class=\"aligncenter size-large\"><img fetchpriority=\"high\" decoding=\"async\" width=\"240\" height=\"319\" src=\"https:\/\/wise.ajou.ac.kr:9605\/wp-content\/uploads\/2021\/06\/yoon2020.jpg\" alt=\"\" class=\"wp-image-853\" srcset=\"https:\/\/wise.ajou.ac.kr\/wp-content\/uploads\/2021\/06\/yoon2020.jpg 240w, https:\/\/wise.ajou.ac.kr\/wp-content\/uploads\/2021\/06\/yoon2020-226x300.jpg 226w\" sizes=\"(max-width: 240px) 100vw, 240px\" \/><\/figure>\n<\/div>\n\n\n<h1 class=\"wp-block-heading has-text-align-center\">\uc724\ub300\uac74<br><sup>Daegun Yoon<\/sup><\/h1>\n\n\n\n<div class=\"wp-block-group has-white-background-color has-background\"><div class=\"wp-block-group__inner-container is-layout-constrained wp-block-group-is-layout-constrained\">\n<div class=\"wp-block-uagb-buttons uagb-buttons__outer-wrap uagb-btn__small-btn uagb-btn-tablet__default-btn uagb-btn-mobile__default-btn uagb-block-cdb053e9\"><div class=\"uagb-buttons__wrap uagb-buttons-layout-wrap\">\n<div class=\"wp-block-uagb-buttons-child uagb-buttons__outer-wrap uagb-block-9ffb071c wp-block-button is-style-outline\"><div class=\"uagb-button__wrapper\"><a class=\"uagb-buttons-repeater wp-block-button__link has-background has-text-color\" href=\"https:\/\/kljp.github.io\/cv.pdf\" onclick=\"return true;\" rel=\"follow noopener\" target=\"_self\"><div class=\"uagb-button__link\">CV<\/div><\/a><\/div><\/div>\n<\/div><\/div>\n<\/div><\/div>\n\n\n\n<h2 class=\"wp-block-heading\">Email<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">kljp at ajou.ac.kr<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Research interests<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Machine Learning, Parallel Algorithm, Distributed System<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Introduction<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Daegun Yoon is a Ph.D. candidate at the Department of Artificial Intelligence, Ajou University. He received the B.S. in the Department of Software, Ajou University, in 2018.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Publications<\/h2>\n\n\n<div class=\"teachpress_pub_list\"><form name=\"tppublistform\" method=\"get\"><a name=\"tppubs\" id=\"tppubs\"><\/a><\/form><div class=\"teachpress_publication_list\"><h3 class=\"tp_h3\" id=\"tp_h3_2024\">2024<\/h3><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">23.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('179','tp_links')\" style=\"cursor:pointer;\">Preserving Near-Optimal Gradient Sparsification Cost for Scalable Distributed Deep Learning<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">The 24th IEEE\/ACM international Symposium on Cluster, Cloud and Internet Computing (CCGrid 2024), <\/span><span class=\"tp_pub_additional_year\">2024<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_resource_link\"><a id=\"tp_links_sh_179\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('179','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_179\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('179','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_179\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{icpp2024yoon,<br \/>\r\ntitle = {Preserving Near-Optimal Gradient Sparsification Cost for Scalable Distributed Deep Learning},<br \/>\r\nauthor = {Daegun Yoon and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/arxiv.org\/abs\/2402.13781},<br \/>\r\nyear  = {2024},<br \/>\r\ndate = {2024-02-13},<br \/>\r\nurldate = {2024-02-13},<br \/>\r\nbooktitle = {The 24th IEEE\/ACM international Symposium on Cluster, Cloud and Internet Computing (CCGrid 2024)},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('179','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_179\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"ai ai-arxiv\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/arxiv.org\/abs\/2402.13781\" title=\"https:\/\/arxiv.org\/abs\/2402.13781\" target=\"_blank\">https:\/\/arxiv.org\/abs\/2402.13781<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('179','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><h3 class=\"tp_h3\" id=\"tp_h3_2023\">2023<\/h3><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">22.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('177','tp_links')\" style=\"cursor:pointer;\">MiCRO: Near-Zero Cost Gradient Sparsification for Scaling and Accelerating Distributed DNN Training<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">30th IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC 2023), <\/span><span class=\"tp_pub_additional_year\">2023<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_resource_link\"><a id=\"tp_links_sh_177\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('177','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_177\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('177','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_177\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{nokey,<br \/>\r\ntitle = {MiCRO: Near-Zero Cost Gradient Sparsification for Scaling and Accelerating Distributed DNN Training},<br \/>\r\nauthor = {Daegun Yoon and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/ieeexplore.ieee.org\/abstract\/document\/10487098},<br \/>\r\nyear  = {2023},<br \/>\r\ndate = {2023-10-02},<br \/>\r\nurldate = {2023-10-02},<br \/>\r\nbooktitle = {30th IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC 2023)},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('177','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_177\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/ieeexplore.ieee.org\/abstract\/document\/10487098\" title=\"https:\/\/ieeexplore.ieee.org\/abstract\/document\/10487098\" target=\"_blank\">https:\/\/ieeexplore.ieee.org\/abstract\/document\/10487098<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('177','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">21.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('176','tp_links')\" style=\"cursor:pointer;\">DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">International Conference on Parallel Processing (ICPP) 2023, <\/span><span class=\"tp_pub_additional_year\">2023<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_176\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('176','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_176\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('176','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_176\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('176','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_176\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{nokey,<br \/>\r\ntitle = {DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification},<br \/>\r\nauthor = {Daegun Yoon and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/dl.acm.org\/doi\/10.1145\/3605573.3605609},<br \/>\r\nyear  = {2023},<br \/>\r\ndate = {2023-08-07},<br \/>\r\nurldate = {2023-08-07},<br \/>\r\nbooktitle = {International Conference on Parallel Processing (ICPP) 2023},<br \/>\r\nabstract = {Gradient sparsification is a widely adopted solution for reducing<br \/>\r\nthe excessive communication traffic in distributed deep learning.<br \/>\r\nHowever, most existing gradient sparsifiers have relatively poor<br \/>\r\nscalability because of considerable computational cost of gradient<br \/>\r\nselection and\/or increased communication traffic owing to gradient<br \/>\r\nbuild-up. To address these challenges, we propose a novel gradient<br \/>\r\nsparsification scheme, DEFT, that partitions the gradient selection<br \/>\r\ntask into sub tasks and distributes them to workers. DEFT differs<br \/>\r\nfrom existing sparsifiers, wherein every worker selects gradients<br \/>\r\namong all gradients. Consequently, the computational cost can<br \/>\r\nbe reduced as the number of workers increases. Moreover, gradi\u0002ent build-up can be eliminated because DEFT allows workers to<br \/>\r\nselect gradients in partitions that are non-intersecting (between<br \/>\r\nworkers). Therefore, even if the number of workers increases, the<br \/>\r\ncommunication traffic can be maintained as per user requirement.<br \/>\r\nTo avoid the loss of significance of gradient selection, DEFT<br \/>\r\nselects more gradients in the layers that have a larger gradient<br \/>\r\nnorm than the other layers. Because every layer has a different<br \/>\r\ncomputational load, DEFT allocates layers to workers using a bin\u0002packing algorithm to maintain a balanced load of gradient selection<br \/>\r\nbetween workers. In our empirical evaluation, DEFT shows a sig\u0002nificant improvement in training performance in terms of speed<br \/>\r\nin gradient selection over existing sparsifiers while achieving high<br \/>\r\nconvergence performance.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('176','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_176\" style=\"display:none;\"><div class=\"tp_abstract_entry\">Gradient sparsification is a widely adopted solution for reducing<br \/>\r\nthe excessive communication traffic in distributed deep learning.<br \/>\r\nHowever, most existing gradient sparsifiers have relatively poor<br \/>\r\nscalability because of considerable computational cost of gradient<br \/>\r\nselection and\/or increased communication traffic owing to gradient<br \/>\r\nbuild-up. To address these challenges, we propose a novel gradient<br \/>\r\nsparsification scheme, DEFT, that partitions the gradient selection<br \/>\r\ntask into sub tasks and distributes them to workers. DEFT differs<br \/>\r\nfrom existing sparsifiers, wherein every worker selects gradients<br \/>\r\namong all gradients. Consequently, the computational cost can<br \/>\r\nbe reduced as the number of workers increases. Moreover, gradi\u0002ent build-up can be eliminated because DEFT allows workers to<br \/>\r\nselect gradients in partitions that are non-intersecting (between<br \/>\r\nworkers). Therefore, even if the number of workers increases, the<br \/>\r\ncommunication traffic can be maintained as per user requirement.<br \/>\r\nTo avoid the loss of significance of gradient selection, DEFT<br \/>\r\nselects more gradients in the layers that have a larger gradient<br \/>\r\nnorm than the other layers. Because every layer has a different<br \/>\r\ncomputational load, DEFT allocates layers to workers using a bin\u0002packing algorithm to maintain a balanced load of gradient selection<br \/>\r\nbetween workers. In our empirical evaluation, DEFT shows a sig\u0002nificant improvement in training performance in terms of speed<br \/>\r\nin gradient selection over existing sparsifiers while achieving high<br \/>\r\nconvergence performance.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('176','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_176\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dl.acm.org\/doi\/10.1145\/3605573.3605609\" title=\"https:\/\/dl.acm.org\/doi\/10.1145\/3605573.3605609\" target=\"_blank\">https:\/\/dl.acm.org\/doi\/10.1145\/3605573.3605609<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('176','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">20.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \uc720\ubbf8\ub9ac,;  \uc724\ub300\uac74,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('173','tp_links')\" style=\"cursor:pointer;\">\uc5f0\ud569\ud559\uc2b5 \uae30\ubc95\ub4e4\uc758 \uc131\ub2a5\ud3c9\uac00\ub97c \uc9c0\uc6d0\ud558\ub294 \uc774\uae30\uc885 \uae30\ubc18\uc758 \uc2e4\ud5d8 \ud50c\ub7ab\ud3fc \uc124\uacc4<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">2023\ub144\ub3c4 \ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ud558\uacc4\uc885\ud569\ud559\uc220\ubc1c\ud45c\ud68c , <\/span><span class=\"tp_pub_additional_organization\"> \ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c <\/span><span class=\"tp_pub_additional_year\">2023<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_resource_link\"><a id=\"tp_links_sh_173\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('173','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_173\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('173','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_173\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{\uc5f0\ud569\ud559\uc2b5\uae30\ubc95\ub4e4\uc758\uc131\ub2a5\ud3c9\uac00\ub97c\uc9c0\uc6d0\ud558\ub294\uc774\uae30\uc885\uae30\ubc18\uc758\uc2e4\ud5d8\ud50c\ub7ab\ud3fc\uc124\uacc4,<br \/>\r\ntitle = {\uc5f0\ud569\ud559\uc2b5 \uae30\ubc95\ub4e4\uc758 \uc131\ub2a5\ud3c9\uac00\ub97c \uc9c0\uc6d0\ud558\ub294 \uc774\uae30\uc885 \uae30\ubc18\uc758 \uc2e4\ud5d8 \ud50c\ub7ab\ud3fc \uc124\uacc4},<br \/>\r\nauthor = {\uc720\ubbf8\ub9ac and \uc724\ub300\uac74 and \uc624\uc0c1\uc724},<br \/>\r\nurl = {https:\/\/www.dbpia.co.kr\/journal\/articleDetail?nodeId=NODE11487802},<br \/>\r\nyear  = {2023},<br \/>\r\ndate = {2023-06-21},<br \/>\r\nurldate = {2023-06-21},<br \/>\r\nbooktitle = {2023\ub144\ub3c4 \ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ud558\uacc4\uc885\ud569\ud559\uc220\ubc1c\ud45c\ud68c },<br \/>\r\norganization = { \ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('173','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_173\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.dbpia.co.kr\/journal\/articleDetail?nodeId=NODE11487802\" title=\"https:\/\/www.dbpia.co.kr\/journal\/articleDetail?nodeId=NODE11487802\" target=\"_blank\">https:\/\/www.dbpia.co.kr\/journal\/articleDetail?nodeId=NODE11487802<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('173','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">19.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Lee, Seungjun;  Yu, Miri;  Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('172','tp_links')\" style=\"cursor:pointer;\">Can hierarchical client clustering mitigate the data heterogeneity effect in federated learning?<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">2023 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), <\/span><span class=\"tp_pub_additional_year\">2023<\/span>, <span class=\"tp_pub_additional_isbn\">ISBN: 979-8-3503-1200-3<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_172\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('172','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_172\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('172','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_172\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('172','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_172\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{nokey,<br \/>\r\ntitle = {Can hierarchical client clustering mitigate the data heterogeneity effect in federated learning?},<br \/>\r\nauthor = {Seungjun Lee and Miri Yu and Daegun Yoon and Sangyoon Oh},<br \/>\r\nurl = {10.1109\/IPDPSW59300.2023.00134},<br \/>\r\ndoi = {10.1109\/IPDPSW59300.2023.00134},<br \/>\r\nisbn = {979-8-3503-1200-3},<br \/>\r\nyear  = {2023},<br \/>\r\ndate = {2023-05-15},<br \/>\r\nurldate = {2023-05-15},<br \/>\r\nbooktitle = {2023 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)},<br \/>\r\nabstract = {Federated learning (FL) was proposed for training a deep neural network model using millions of user data. The technique has attracted considerable attention owing to its privacy-preserving characteristic. However, two major challenges exist. The first is the limitation of simultaneously participating clients. If the number of clients increases, the single parameter server easily becomes a bottleneck and is prone to have stragglers. The second is data heterogeneity, which adversely affects the accuracy of the global model. Because data should remain at user devices to preserve privacy, we cannot use data shuffling, which is used to homogenize training data in traditional distributed deep learning. We propose a client clustering and model aggregation method, CCFed, to increase the number of simultaneously participating clients and mitigate the data heterogeneity problem. CCFed improves the learning performance using set partition modeling to let data be evenly distributed between clusters and mitigate the effect of a non-IID environment. Experiments show that we can achieve a 2.7-14% higher accuracy using CCFed compared with FedAvg, where CCFed requires approximately 50% less number of rounds compared with FedAvg training on benchmark datasets.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('172','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_172\" style=\"display:none;\"><div class=\"tp_abstract_entry\">Federated learning (FL) was proposed for training a deep neural network model using millions of user data. The technique has attracted considerable attention owing to its privacy-preserving characteristic. However, two major challenges exist. The first is the limitation of simultaneously participating clients. If the number of clients increases, the single parameter server easily becomes a bottleneck and is prone to have stragglers. The second is data heterogeneity, which adversely affects the accuracy of the global model. Because data should remain at user devices to preserve privacy, we cannot use data shuffling, which is used to homogenize training data in traditional distributed deep learning. We propose a client clustering and model aggregation method, CCFed, to increase the number of simultaneously participating clients and mitigate the data heterogeneity problem. CCFed improves the learning performance using set partition modeling to let data be evenly distributed between clusters and mitigate the effect of a non-IID environment. Experiments show that we can achieve a 2.7-14% higher accuracy using CCFed compared with FedAvg, where CCFed requires approximately 50% less number of rounds compared with FedAvg training on benchmark datasets.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('172','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_172\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"10.1109\/IPDPSW59300.2023.00134\" title=\"10.1109\/IPDPSW59300.2023.00134\" target=\"_blank\">10.1109\/IPDPSW59300.2023.00134<\/a><\/li><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/10.1109\/IPDPSW59300.2023.00134\" title=\"Follow DOI:10.1109\/IPDPSW59300.2023.00134\" target=\"_blank\">doi:10.1109\/IPDPSW59300.2023.00134<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('172','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">18.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \ucd5c\uc9c0\ud5cc,;  \uc720\ubbf8\ub9ac,;  \uc724\ub300\uac74,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('168','tp_links')\" style=\"cursor:pointer;\">\uc5f0\ud569\ud559\uc2b5\uc5d0\uc11c\uc758 \ubcf4\uc548 \ucde8\uc57d\uc810 \ubd84\uc11d<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">2023\ub144\ub3c4 \ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ub3d9\uacc4\uc885\ud569\ud559\uc220\ubc1c\ud45c\ud68c \ub17c\ubb38\uc9d1\r\n, <\/span><span class=\"tp_pub_additional_volume\">vol. 80, <\/span><span class=\"tp_pub_additional_organization\">\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c <\/span><span class=\"tp_pub_additional_year\">2023<\/span>, <span class=\"tp_pub_additional_issn\">ISSN: 2383-8302<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_168\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('168','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_168\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('168','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_168\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('168','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_168\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{\ucd5c\uc9c0\ud5cc2023\uc5f0\ud569\ud559\uc2b5\uc5d0\uc11c\uc758,<br \/>\r\ntitle = {\uc5f0\ud569\ud559\uc2b5\uc5d0\uc11c\uc758 \ubcf4\uc548 \ucde8\uc57d\uc810 \ubd84\uc11d},<br \/>\r\nauthor = {\ucd5c\uc9c0\ud5cc and \uc720\ubbf8\ub9ac and \uc724\ub300\uac74 and \uc624\uc0c1\uc724},<br \/>\r\nurl = {https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE11227811},<br \/>\r\nissn = {2383-8302},<br \/>\r\nyear  = {2023},<br \/>\r\ndate = {2023-02-28},<br \/>\r\nurldate = {2023-02-28},<br \/>\r\nbooktitle = {2023\ub144\ub3c4 \ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ub3d9\uacc4\uc885\ud569\ud559\uc220\ubc1c\ud45c\ud68c \ub17c\ubb38\uc9d1<br \/>\r\n},<br \/>\r\nvolume = {80},<br \/>\r\npages = {1201-1202},<br \/>\r\norganization = {\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c},<br \/>\r\nabstract = {\uac1c\uc778 \ub370\uc774\ud130\uc5d0 \ub300\ud55c \ud504\ub77c\uc774\ubc84\uc2dc \uce68\ud574 \uc5c6\uc774 \ubd84\uc0b0 \uae30\uacc4\ud559\uc2b5\uc744 \uad6c\ud604\ud558\uae30 \uc704\ud574 \uc5f0\ud569\ud559\uc2b5\uc774 \uc81c\uc548\ub418\uc5c8\ub2e4. \uae30\uc874 \uc5f0\ud569\ud559\uc2b5 \uae30\ubc95\uc758 \uac1c\uc120\uc744 \ud1b5\ud574 \uc815\ud655\ub3c4\ud5a5\uc0c1 \ubc0f \uc218\ub834\uc18d\ub3c4 \ud5a5\uc0c1\uc744 \ubaa9\ud45c\ub85c \ud558\ub294 \uc0c8\ub85c\uc6b4 \uae30\ubc95\ub4e4\uc774 \ub4f1\uc7a5\ud558\uace0 \uc788\uc5b4\uc11c, \uc774\uc5d0 \ub300\ud55c \ubcf4\uc548 \uac00\uc774\ub4dc\ub77c\uc778\uc774 \ud544\uc694\ud55c \uc0c1\ud669\uc774\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294\uc5f0\ud569\ud559\uc2b5 \uad6c\uc870\uc758 \ud2b9\uc9d5\uc73c\ub85c \ub098\ud0c0\ub098\ub294 \ubcf4\uc548 \ucde8\uc57d\uc810\uc744 \uacf5\uaca9\ud615\ud0dc \ubcc4\ub85c \uad6c\ubd84\ud558\uace0 \uc774\uc5d0 \ub300\ud55c \ub300\uc751\ubc29\uc548\uc744 \uace0\ucc30\ud55c\ub2e4.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('168','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_168\" style=\"display:none;\"><div class=\"tp_abstract_entry\">\uac1c\uc778 \ub370\uc774\ud130\uc5d0 \ub300\ud55c \ud504\ub77c\uc774\ubc84\uc2dc \uce68\ud574 \uc5c6\uc774 \ubd84\uc0b0 \uae30\uacc4\ud559\uc2b5\uc744 \uad6c\ud604\ud558\uae30 \uc704\ud574 \uc5f0\ud569\ud559\uc2b5\uc774 \uc81c\uc548\ub418\uc5c8\ub2e4. \uae30\uc874 \uc5f0\ud569\ud559\uc2b5 \uae30\ubc95\uc758 \uac1c\uc120\uc744 \ud1b5\ud574 \uc815\ud655\ub3c4\ud5a5\uc0c1 \ubc0f \uc218\ub834\uc18d\ub3c4 \ud5a5\uc0c1\uc744 \ubaa9\ud45c\ub85c \ud558\ub294 \uc0c8\ub85c\uc6b4 \uae30\ubc95\ub4e4\uc774 \ub4f1\uc7a5\ud558\uace0 \uc788\uc5b4\uc11c, \uc774\uc5d0 \ub300\ud55c \ubcf4\uc548 \uac00\uc774\ub4dc\ub77c\uc778\uc774 \ud544\uc694\ud55c \uc0c1\ud669\uc774\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294\uc5f0\ud569\ud559\uc2b5 \uad6c\uc870\uc758 \ud2b9\uc9d5\uc73c\ub85c \ub098\ud0c0\ub098\ub294 \ubcf4\uc548 \ucde8\uc57d\uc810\uc744 \uacf5\uaca9\ud615\ud0dc \ubcc4\ub85c \uad6c\ubd84\ud558\uace0 \uc774\uc5d0 \ub300\ud55c \ub300\uc751\ubc29\uc548\uc744 \uace0\ucc30\ud55c\ub2e4.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('168','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_168\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE11227811\" title=\"https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE11227811\" target=\"_blank\">https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE11227811<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('168','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">17.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Jeong, Minjoong;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('167','tp_links')\" style=\"cursor:pointer;\">SAGE: toward on-the-fly gradient compression ratio scaling<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">The Journal of Supercomputing, <\/span><span class=\"tp_pub_additional_pages\">pp. 1\u201323, <\/span><span class=\"tp_pub_additional_year\">2023<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_167\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('167','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_167\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('167','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_167\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('167','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_167\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{yoon2023sage,<br \/>\r\ntitle = {SAGE: toward on-the-fly gradient compression ratio scaling},<br \/>\r\nauthor = {Daegun Yoon and Minjoong Jeong and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/link.springer.com\/article\/10.1007\/s11227-023-05120-7},<br \/>\r\ndoi = {https:\/\/doi.org\/10.1007\/s11227-023-05120-7},<br \/>\r\nyear  = {2023},<br \/>\r\ndate = {2023-02-25},<br \/>\r\nurldate = {2023-02-25},<br \/>\r\njournal = {The Journal of Supercomputing},<br \/>\r\npages = {1--23},<br \/>\r\nabstract = {Gradient sparsification is widely adopted in distributed training; however, it suffers from a trade-off between computation and communication. The prevalent Top-k sparsifier has a hard constraint on computational overhead while achieving the desired gradient compression ratio. Conversely, the hard-threshold sparsifier eliminates computational constraints but fail to achieve the targeted compression ratio. Motivated by this tradeoff, we designed a novel threshold-based sparsifier called SAGE, which achieves a compression ratio close to that of the Top-k sparsifier with negligible computational overhead. SAGE scales the compression ratio by deriving an adjustable threshold based on each iteration\u2019s heuristics. Experimental results show that SAGE achieves a compression ratio closer to the desired ratio than a hard-threshold sparsifier without exacerbating the accuracy of model training. In terms of computation time for gradient selection, SAGE achieves a speedup of up to 23.62\u00d7<br \/>\r\n over the Top-k sparsifier.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('167','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_167\" style=\"display:none;\"><div class=\"tp_abstract_entry\">Gradient sparsification is widely adopted in distributed training; however, it suffers from a trade-off between computation and communication. The prevalent Top-k sparsifier has a hard constraint on computational overhead while achieving the desired gradient compression ratio. Conversely, the hard-threshold sparsifier eliminates computational constraints but fail to achieve the targeted compression ratio. Motivated by this tradeoff, we designed a novel threshold-based sparsifier called SAGE, which achieves a compression ratio close to that of the Top-k sparsifier with negligible computational overhead. SAGE scales the compression ratio by deriving an adjustable threshold based on each iteration\u2019s heuristics. Experimental results show that SAGE achieves a compression ratio closer to the desired ratio than a hard-threshold sparsifier without exacerbating the accuracy of model training. In terms of computation time for gradient selection, SAGE achieves a speedup of up to 23.62\u00d7<br \/>\r\n over the Top-k sparsifier.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('167','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_167\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/link.springer.com\/article\/10.1007\/s11227-023-05120-7\" title=\"https:\/\/link.springer.com\/article\/10.1007\/s11227-023-05120-7\" target=\"_blank\">https:\/\/link.springer.com\/article\/10.1007\/s11227-023-05120-7<\/a><\/li><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/https:\/\/doi.org\/10.1007\/s11227-023-05120-7\" title=\"Follow DOI:https:\/\/doi.org\/10.1007\/s11227-023-05120-7\" target=\"_blank\">doi:https:\/\/doi.org\/10.1007\/s11227-023-05120-7<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('167','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><h3 class=\"tp_h3\" id=\"tp_h3_2022\">2022<\/h3><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">16.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Jeong, Minjoong;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('165','tp_links')\" style=\"cursor:pointer;\">WAVE: designing a heuristics-based three-way breadth-first search on GPUs<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">The Journal of Supercomputing, <\/span><span class=\"tp_pub_additional_year\">2022<\/span><span class=\"tp_pub_additional_note\">, (2)<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_165\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('165','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_165\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('165','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_165\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('165','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_165\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{Yoon2022WAVE,<br \/>\r\ntitle = {WAVE: designing a heuristics-based three-way breadth-first search on GPUs},<br \/>\r\nauthor = {Daegun Yoon and Minjoong Jeong and Sangyoon Oh},<br \/>\r\ndoi = {10.1007\/s11227-022-04934-1},<br \/>\r\nyear  = {2022},<br \/>\r\ndate = {2022-11-17},<br \/>\r\nurldate = {2022-11-17},<br \/>\r\njournal = {The Journal of Supercomputing},<br \/>\r\nabstract = {Breadth-first search (BFS) is a building block for improving the performance of many iterative graph algorithms. In addition to conventional BFS (push), a novel method that traverses a graph in the reverse direction (pull) has emerged and gained popularity because of its enhanced processing performance. Several frameworks have recently used a hybrid approach known as direction-optimizing BFS, which utilizes both directions. However, these frameworks are mostly interested in optimizing the procedure in each direction, instead of designing sophisticated methods for determining the appropriate direction between push and pull at each iteration. Owing to the lack of in-depth discussion on this decision, state-of-the-art direction-optimizing BFS algorithms cannot demonstrate their comprehensive performance despite improvements in the design of each direction because they select ineffective directions at each iteration. We identified that the current frameworks suffer from high computational overheads for their decisions and make decisions that are overfitted to several graph datasets used for tuning their direction selection process. Based on observations from state-of-the-art limitations, we designed a direction-optimizing method for BFS called WAVE. WAVE minimizes the computational overhead to near zero and makes more appropriate direction selection decisions than the state-of-the-art heuristics based on the characteristics extracted from the input graph dataset. In our experiments on 20 graph benchmarks, WAVE achieved speedups of up to 4.95\u00d7, 5.79\u00d7, 46.49\u00d7, and 149.67\u00d7 over Enterprise, Gunrock, Tigr, and CuSha, respectively.},<br \/>\r\nnote = {2},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('165','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_165\" style=\"display:none;\"><div class=\"tp_abstract_entry\">Breadth-first search (BFS) is a building block for improving the performance of many iterative graph algorithms. In addition to conventional BFS (push), a novel method that traverses a graph in the reverse direction (pull) has emerged and gained popularity because of its enhanced processing performance. Several frameworks have recently used a hybrid approach known as direction-optimizing BFS, which utilizes both directions. However, these frameworks are mostly interested in optimizing the procedure in each direction, instead of designing sophisticated methods for determining the appropriate direction between push and pull at each iteration. Owing to the lack of in-depth discussion on this decision, state-of-the-art direction-optimizing BFS algorithms cannot demonstrate their comprehensive performance despite improvements in the design of each direction because they select ineffective directions at each iteration. We identified that the current frameworks suffer from high computational overheads for their decisions and make decisions that are overfitted to several graph datasets used for tuning their direction selection process. Based on observations from state-of-the-art limitations, we designed a direction-optimizing method for BFS called WAVE. WAVE minimizes the computational overhead to near zero and makes more appropriate direction selection decisions than the state-of-the-art heuristics based on the characteristics extracted from the input graph dataset. In our experiments on 20 graph benchmarks, WAVE achieved speedups of up to 4.95\u00d7, 5.79\u00d7, 46.49\u00d7, and 149.67\u00d7 over Enterprise, Gunrock, Tigr, and CuSha, respectively.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('165','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_165\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/10.1007\/s11227-022-04934-1\" title=\"Follow DOI:10.1007\/s11227-022-04934-1\" target=\"_blank\">doi:10.1007\/s11227-022-04934-1<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('165','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">15.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \uc724\ub300\uac74,;  \ub178\ubcd1\ud76c,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('163','tp_links')\" style=\"cursor:pointer;\">\uc804\uc220\ub9dd\uc758 \ub77c\uc6b0\ud305 \uc131\ub2a5 \uac1c\uc120\uc744 \uc704\ud55c \uc131\ub2a5 \uc9c0\ud45c \ubd84\uc11d \uae30\ubc18 \uc815\ucc45 \uc5d4\uc9c4 \uc124\uacc4<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ub17c\ubb38\uc9c0, <\/span><span class=\"tp_pub_additional_volume\">vol. 47, <\/span><span class=\"tp_pub_additional_issue\">iss. 9, <\/span><span class=\"tp_pub_additional_number\">no. 9, <\/span><span class=\"tp_pub_additional_pages\">pp. 1353-1359, <\/span><span class=\"tp_pub_additional_year\">2022<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_163\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('163','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_163\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('163','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_163\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('163','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_163\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{\uc724\ub300\uac742022\uc804\uc220\ub9dd,<br \/>\r\ntitle = {\uc804\uc220\ub9dd\uc758 \ub77c\uc6b0\ud305 \uc131\ub2a5 \uac1c\uc120\uc744 \uc704\ud55c \uc131\ub2a5 \uc9c0\ud45c \ubd84\uc11d \uae30\ubc18 \uc815\ucc45 \uc5d4\uc9c4 \uc124\uacc4},<br \/>\r\nauthor = {\uc724\ub300\uac74 and \ub178\ubcd1\ud76c and \uc624\uc0c1\uc724},<br \/>\r\nurl = {https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART002877193},<br \/>\r\nyear  = {2022},<br \/>\r\ndate = {2022-10-31},<br \/>\r\nurldate = {2022-10-31},<br \/>\r\njournal = {\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ub17c\ubb38\uc9c0},<br \/>\r\nvolume = {47},<br \/>\r\nnumber = {9},<br \/>\r\nissue = {9},<br \/>\r\npages = {1353-1359},<br \/>\r\nabstract = {\ucef4\ud4e8\ud305 \uad00\ub828 \uae30\uc220 \ubc1c\ub2ec\uc5d0 \ub530\ub77c \uad70 \uc791\uc804 \uc218\ud589\uc5d0\uc11c \ubc1c\uc0dd\ud558\ub294 \ub370\uc774\ud130\uc758 \uaddc\ubaa8\uac00 \ub9e4\uc6b0 \ucee4\uc9c0\uace0 \uc788\uc73c\uba70, \uc774\uc5d0 \ub530\ub77c \uc774\ub97c \ucc98\ub9ac\ud558\uae30 \uc704\ud55c \uad70 \uc804\uc220\ub9dd\uc758 \uc131\ub2a5 \ud5a5\uc0c1\uc5d0 \ub300\ud55c \uc694\uad6c \ub610\ud55c \uc810\uc810 \ub298\uc5b4\ub098\uace0 \uc788\ub2e4. \uad70 \uc804\uc220\ub9dd\uc758 \ud2b9\uc131 \uc0c1 \ub2e4\uc591\ud55c \uc7a5\ube44\ub85c \uad6c\uc131\ub41c \ub124\ud2b8\uc6cc\ud06c\ub97c \ud65c\uc6a9\ud574\uc57c \ud558\uba70, \uc774\ub7ec\ud55c \uc0c1\ud669\uc5d0\uc11c \ubbfc\uac04\uc5d0\uc11c \ud65c\ubc1c\ud788 \uc801\uc6a9\ub418\ub294 Software-Defined Network (SDN) \uae30\uc220\uc744 \uc801\uc6a9\ud55c\ub2e4\uba74 \uc7a5\ube44\ub97c \uc81c\uacf5 \ubca4\ub354\ub85c\ubd80\ud130 \uc790\uc720\ub85c\uc6b4 \uc190\uc26c\uc6b4 \ub124\ud2b8\uc6cc\ud06c \uad00\ub9ac\uac00 \uac00\ub2a5\ud558\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294SDN \uae30\ubc18 \ub124\ud2b8\uc6cc\ud06c \ud658\uacbd\uc5d0\uc11c \ud328\ud0b7 \uc804\uc1a1 \uc131\ub2a5 \ud5a5\uc0c1\uc744 \ubaa9\uc801\uc73c\ub85c \ud558\ub294 \ub124\ud2b8\uc6cc\ud06c \uc815\ucc45 \uc5d4\uc9c4 \uad6c\uc870 \uc124\uacc4\ub97c \uc18c\uac1c\ud55c\ub2e4.<br \/>\r\n\uc815\ucc45 \uc5d4\uc9c4\uc740 Flow table\uc758 Flow\ub4e4\uc774 \ub098\ud0c0\ub0b4\ub294 \ub77c\uc6b0\ud305 \uacbd\ub85c\ub97c \uc218\uc815\ud558\ub3c4\ub85d \ud558\ub294 \uc54c\uace0\ub9ac\uc998\uc744 \ud3ec\ud568\ud558\uba70 \uc131\ub2a5 \uac1c\uc120 \uc5ec\ubd80\ub294 \ubcf8 \uc5f0\uad6c\uc5d0\uc11c \uc815\uc758\ud55c \uc885\ud569 \uc131\ub2a5 \uc9c0\ud45c\ub97c \ud1b5\ud574 \ud310\ub2e8\ud55c\ub2e4. \ucd94\ud6c4 \ubcf8 \uc5f0\uad6c\uc5d0\uc11c \uc81c\uc548\ud558\ub294 \uc804\uc220\ub9dd \ub77c\uc6b0\ud305 \uc131\ub2a5 \uac1c\ub7c9\uc744 \uc704\ud55c \uc131\ub2a5 \uc9c0\ud45c \ubd84\uc11d \uae30\ubc18 \uc815\ucc45 \uc5d4\uc9c4 \uae30\ubc18\uc758 \uc18c\ud504\ud2b8\uc6e8\uc5b4\ub97c \uc2e4\uc81c \ub124\ud2b8\uc6cc\ud06c \uc6b4\uc6a9 \uc0c1\ud669\uc5d0 \uc801\uc6a9\ud558\uace0, \ub124\ud2b8\uc6cc\ud06c \uc131\ub2a5 \ud5a5\uc0c1\uc744 \uac80\uc99d\ud558\ub3c4\ub85d \ud560 \uacc4\ud68d\uc774\ub2e4.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('163','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_163\" style=\"display:none;\"><div class=\"tp_abstract_entry\">\ucef4\ud4e8\ud305 \uad00\ub828 \uae30\uc220 \ubc1c\ub2ec\uc5d0 \ub530\ub77c \uad70 \uc791\uc804 \uc218\ud589\uc5d0\uc11c \ubc1c\uc0dd\ud558\ub294 \ub370\uc774\ud130\uc758 \uaddc\ubaa8\uac00 \ub9e4\uc6b0 \ucee4\uc9c0\uace0 \uc788\uc73c\uba70, \uc774\uc5d0 \ub530\ub77c \uc774\ub97c \ucc98\ub9ac\ud558\uae30 \uc704\ud55c \uad70 \uc804\uc220\ub9dd\uc758 \uc131\ub2a5 \ud5a5\uc0c1\uc5d0 \ub300\ud55c \uc694\uad6c \ub610\ud55c \uc810\uc810 \ub298\uc5b4\ub098\uace0 \uc788\ub2e4. \uad70 \uc804\uc220\ub9dd\uc758 \ud2b9\uc131 \uc0c1 \ub2e4\uc591\ud55c \uc7a5\ube44\ub85c \uad6c\uc131\ub41c \ub124\ud2b8\uc6cc\ud06c\ub97c \ud65c\uc6a9\ud574\uc57c \ud558\uba70, \uc774\ub7ec\ud55c \uc0c1\ud669\uc5d0\uc11c \ubbfc\uac04\uc5d0\uc11c \ud65c\ubc1c\ud788 \uc801\uc6a9\ub418\ub294 Software-Defined Network (SDN) \uae30\uc220\uc744 \uc801\uc6a9\ud55c\ub2e4\uba74 \uc7a5\ube44\ub97c \uc81c\uacf5 \ubca4\ub354\ub85c\ubd80\ud130 \uc790\uc720\ub85c\uc6b4 \uc190\uc26c\uc6b4 \ub124\ud2b8\uc6cc\ud06c \uad00\ub9ac\uac00 \uac00\ub2a5\ud558\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294SDN \uae30\ubc18 \ub124\ud2b8\uc6cc\ud06c \ud658\uacbd\uc5d0\uc11c \ud328\ud0b7 \uc804\uc1a1 \uc131\ub2a5 \ud5a5\uc0c1\uc744 \ubaa9\uc801\uc73c\ub85c \ud558\ub294 \ub124\ud2b8\uc6cc\ud06c \uc815\ucc45 \uc5d4\uc9c4 \uad6c\uc870 \uc124\uacc4\ub97c \uc18c\uac1c\ud55c\ub2e4.<br \/>\r\n\uc815\ucc45 \uc5d4\uc9c4\uc740 Flow table\uc758 Flow\ub4e4\uc774 \ub098\ud0c0\ub0b4\ub294 \ub77c\uc6b0\ud305 \uacbd\ub85c\ub97c \uc218\uc815\ud558\ub3c4\ub85d \ud558\ub294 \uc54c\uace0\ub9ac\uc998\uc744 \ud3ec\ud568\ud558\uba70 \uc131\ub2a5 \uac1c\uc120 \uc5ec\ubd80\ub294 \ubcf8 \uc5f0\uad6c\uc5d0\uc11c \uc815\uc758\ud55c \uc885\ud569 \uc131\ub2a5 \uc9c0\ud45c\ub97c \ud1b5\ud574 \ud310\ub2e8\ud55c\ub2e4. \ucd94\ud6c4 \ubcf8 \uc5f0\uad6c\uc5d0\uc11c \uc81c\uc548\ud558\ub294 \uc804\uc220\ub9dd \ub77c\uc6b0\ud305 \uc131\ub2a5 \uac1c\ub7c9\uc744 \uc704\ud55c \uc131\ub2a5 \uc9c0\ud45c \ubd84\uc11d \uae30\ubc18 \uc815\ucc45 \uc5d4\uc9c4 \uae30\ubc18\uc758 \uc18c\ud504\ud2b8\uc6e8\uc5b4\ub97c \uc2e4\uc81c \ub124\ud2b8\uc6cc\ud06c \uc6b4\uc6a9 \uc0c1\ud669\uc5d0 \uc801\uc6a9\ud558\uace0, \ub124\ud2b8\uc6cc\ud06c \uc131\ub2a5 \ud5a5\uc0c1\uc744 \uac80\uc99d\ud558\ub3c4\ub85d \ud560 \uacc4\ud68d\uc774\ub2e4.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('163','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_163\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART002877193\" title=\"https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArti[...]\" target=\"_blank\">https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArti[...]<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('163','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">14.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('166','tp_links')\" style=\"cursor:pointer;\">Empirical Analysis on Top-k Gradient Sparsification for Distributed Deep Learning in a Supercomputing Environment<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">The 8th International Conference on Next Generation Computing (ICNGC) 2022, <\/span><span class=\"tp_pub_additional_year\">2022<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_166\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('166','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_166\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('166','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_166\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('166','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_166\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{yoon2022empirical,<br \/>\r\ntitle = {Empirical Analysis on Top-k Gradient Sparsification for Distributed Deep Learning in a Supercomputing Environment},<br \/>\r\nauthor = {Daegun Yoon and Sangyoon Oh},<br \/>\r\ndoi = {10.48550\/arXiv.2209.08497},<br \/>\r\nyear  = {2022},<br \/>\r\ndate = {2022-09-19},<br \/>\r\nbooktitle = {The 8th International Conference on Next Generation Computing (ICNGC) 2022},<br \/>\r\nabstract = {To train deep learning models faster, distributed training on multiple GPUs is the very popular scheme in recent years. However, the communication bandwidth is still a major bottleneck of training performance. To improve overall training performance, recent works have proposed gradient sparsification methods that reduce the communication traffic significantly. Most of them require gradient sorting to select meaningful gradients such as Top-k gradient sparsification (Top-k SGD). However, Top-k SGD has a limit to increase the speed up overall training performance because gradient sorting is significantly inefficient on GPUs. In this paper, we conduct experiments that show the inefficiency of Top-k SGD and provide the insight of the low performance. Based on observations from our empirical analysis, we plan to yield a high performance gradient sparsification method as a future work. },<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('166','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_166\" style=\"display:none;\"><div class=\"tp_abstract_entry\">To train deep learning models faster, distributed training on multiple GPUs is the very popular scheme in recent years. However, the communication bandwidth is still a major bottleneck of training performance. To improve overall training performance, recent works have proposed gradient sparsification methods that reduce the communication traffic significantly. Most of them require gradient sorting to select meaningful gradients such as Top-k gradient sparsification (Top-k SGD). However, Top-k SGD has a limit to increase the speed up overall training performance because gradient sorting is significantly inefficient on GPUs. In this paper, we conduct experiments that show the inefficiency of Top-k SGD and provide the insight of the low performance. Based on observations from our empirical analysis, we plan to yield a high performance gradient sparsification method as a future work. <\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('166','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_166\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/10.48550\/arXiv.2209.08497\" title=\"Follow DOI:10.48550\/arXiv.2209.08497\" target=\"_blank\">doi:10.48550\/arXiv.2209.08497<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('166','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">13.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \uc774\uc2b9\uc900,;  \uc724\ub300\uac74,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('160','tp_links')\" style=\"cursor:pointer;\">SDN \uc815\ucc45\uc5d4\uc9c4\uc758 \uc0ac\uc6a9\uc790 \ubaa8\ub4c8\uc744 \uc704\ud55c \ubd84\uc11d \uc694\uccad \uc815\uc758 \uc5b8\uc5b4<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ub17c\ubb38\uc9c0, <\/span><span class=\"tp_pub_additional_volume\">vol. 47, <\/span><span class=\"tp_pub_additional_number\">no. 9, <\/span><span class=\"tp_pub_additional_pages\">pp. 1360-1369, <\/span><span class=\"tp_pub_additional_year\">2022<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_160\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('160','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_160\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('160','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_160\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('160','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_160\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{\uc774\uc2b9\uc9002022SDN,<br \/>\r\ntitle = {SDN \uc815\ucc45\uc5d4\uc9c4\uc758 \uc0ac\uc6a9\uc790 \ubaa8\ub4c8\uc744 \uc704\ud55c \ubd84\uc11d \uc694\uccad \uc815\uc758 \uc5b8\uc5b4},<br \/>\r\nauthor = {\uc774\uc2b9\uc900 and \uc724\ub300\uac74 and \uc624\uc0c1\uc724},<br \/>\r\nurl = {https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART002877194},<br \/>\r\nyear  = {2022},<br \/>\r\ndate = {2022-09-01},<br \/>\r\nurldate = {2022-09-01},<br \/>\r\njournal = {\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ub17c\ubb38\uc9c0},<br \/>\r\nvolume = {47},<br \/>\r\nnumber = {9},<br \/>\r\npages = {1360-1369},<br \/>\r\nabstract = {\ud604\ub300\uc804\uc5d0\uc11c \uc791\uc804 \uc218\ud589\uc740 \ub124\ud2b8\uc6cc\ud06c \uc911\uc2ec\uc804\uc758 \uc591\uc0c1\uc744 \ub744\uace0 \uc788\uc73c\uba70, \uc774\uc5d0 \ub530\ub77c \uad70 \uc804\uc220\ub9dd \uc790\uc6d0\uc744 \ud6a8\uacfc\uc801\uc73c\ub85c \uc0ac\uc6a9\ud558\uae30 \uc704\ud55c \uc5ec\ub7ec \uc5f0\uad6c\uac00 \uc9c4\ud589\ub418\uace0 \uc788\ub2e4. \ub2e8\uc77c \uc5f0\uad6c \uacb0\uacfc\uac00 \uc544\ub2cc \uc5ec\ub7ec \uc5f0\uad6c\uc758 \uacb0\uacfc\ub97c \ubcf5\ud569\uc801\uc73c\ub85c \uc801\uc6a9\ud588\uc744 \ub54c\uc758 \ud6a8\uacfc\ub97c \ubd84\uc11d\ud558\uae30 \uc704\ud55c \ub178\ub825\uc758 \uc77c\ud658\uc73c\ub85c \ud1b5\ud569 \ud14c\uc2a4\ud2b8\ubca0\ub4dc\uac00 \uad6c\ucd95\ub418\uace0, \uc5ec\uae30\uc5d0\uc11c \uc5ec\ub7ec \ub124\ud2b8\uc6cc\ud06c \uc54c\uace0\ub9ac\uc998\uc744 \ub3d9\uc2dc\uc5d0 \uc218\ud589\ud558\uace0 \uc131\ub2a5\uc744 \ubd84\uc11d\ud558\uae30 \uc704\ud55c \uc815\ucc45 \uc5d4\uc9c4\ub3c4 \uc124\uacc4\ub418\uc5c8\ub2e4. \ud558\uc9c0\ub9cc \uc774\uc885 \ub124\ud2b8\uc6cc\ud06c \ud658\uacbd\uc5d0\uc11c\ub294 \uc0ac\uc6a9\uc790\ub4e4\uc774 \uc694\uad6c\ud558\ub294 \uc11c\ub978 \ub2e4\ub978 \ub370\uc774\ud130 \uad6c\uc870\uc758 \uc131\ub2a5 \uc9c0\ud45c\uc640 \uc774\ub97c \ucc98\ub9ac\ud560 \uac01 \uc54c\uace0\ub9ac\uc998\uc758 \uc11c\ub85c \ub2e4\ub978 \uc2e4\ud589 \ud658\uacbd\uc5d0 \uc801\uc751\uc801\uc73c\ub85c \ub300\uc751\ud558\uae30 \uc5b4\ub824\uc6b4 \ubb38\uc81c\uac00 \uc788\uc5c8\ub2e4. \uc774\uc5d0 \ubcf8 \uc5f0\uad6c\uc5d0\uc11c\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4\uc640 \uc2e4\ud589 \ud658\uacbd \ub4f1 \ud2b9\uc815 \uae30\uc220\uc5d0 \uc885\uc18d\ub418\uc9c0 \uc54a\ub294 \uc815\ucc45 \uc5d4\uc9c4\uc744 \uc704\ud55c XML \uae30\ubc18\uc758 \uc778\ud130\ud398\uc774\uc2a4 \ud3ec\ub9f7\uc744 \uc815\uc758\ud558\uace0 \uadf8 \uc2a4\ud0a4\ub9c8\ub97c \uc81c\uc548\ud55c\ub2e4. \uc81c\uc548\ub41c \uc2a4\ud0a4\ub9c8\ub97c \uc0ac\uc6a9\ud558\uc5ec \uba54\uc2dc\uc9c0\ub294 \ud2b9\uc815 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4\uc5d0 \uc885\uc18d\ub418\uc9c0 \uc54a\uace0 \uc778\ucf54\ub529\uacfc \ub514\ucf54\ub529\uc744 \ud560 \uc218 \uc788\uc73c\uba70 Open Container Initiative \ud45c\uc900\uc744 \uae30\ubc18\uc73c\ub85c\uc2e4\ud589 \ud658\uacbd\uc744 \uc815\uc758\ud558\ub294 \ucee8\ud14c\uc774\ub108\ub97c \uae30\uc220\ud560 \uc218 \uc788\ub2e4.<br \/>\r\n<br \/>\r\nIn modern warfare environment, the well defined networks becomes important to the operations. Thus, researchers study on how to use the military tactical network resources effectively. To analyze effectiveness of the results from multiple studies together, an integrated testbed is critical as well as the design and the implementation of a policy engine that performs multiple network algorithms and analyze performance simultaneously. However, when the network environment is heterogeneous, it is hard to respond adaptively to the performance indicators of the different data structures and the different execution environments of each user algorithm. To address this issue, we propose an XML-based interface format and its schema for the policy engine, which is independent from specific technologies such as programming languages and execution environments. A message from and to the policy engine and the testbed can be encoded and decoded regardless of the programming language. Furthermore, it can describe containers of the execution environment based on the Open Container Initiative standard.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('160','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_160\" style=\"display:none;\"><div class=\"tp_abstract_entry\">\ud604\ub300\uc804\uc5d0\uc11c \uc791\uc804 \uc218\ud589\uc740 \ub124\ud2b8\uc6cc\ud06c \uc911\uc2ec\uc804\uc758 \uc591\uc0c1\uc744 \ub744\uace0 \uc788\uc73c\uba70, \uc774\uc5d0 \ub530\ub77c \uad70 \uc804\uc220\ub9dd \uc790\uc6d0\uc744 \ud6a8\uacfc\uc801\uc73c\ub85c \uc0ac\uc6a9\ud558\uae30 \uc704\ud55c \uc5ec\ub7ec \uc5f0\uad6c\uac00 \uc9c4\ud589\ub418\uace0 \uc788\ub2e4. \ub2e8\uc77c \uc5f0\uad6c \uacb0\uacfc\uac00 \uc544\ub2cc \uc5ec\ub7ec \uc5f0\uad6c\uc758 \uacb0\uacfc\ub97c \ubcf5\ud569\uc801\uc73c\ub85c \uc801\uc6a9\ud588\uc744 \ub54c\uc758 \ud6a8\uacfc\ub97c \ubd84\uc11d\ud558\uae30 \uc704\ud55c \ub178\ub825\uc758 \uc77c\ud658\uc73c\ub85c \ud1b5\ud569 \ud14c\uc2a4\ud2b8\ubca0\ub4dc\uac00 \uad6c\ucd95\ub418\uace0, \uc5ec\uae30\uc5d0\uc11c \uc5ec\ub7ec \ub124\ud2b8\uc6cc\ud06c \uc54c\uace0\ub9ac\uc998\uc744 \ub3d9\uc2dc\uc5d0 \uc218\ud589\ud558\uace0 \uc131\ub2a5\uc744 \ubd84\uc11d\ud558\uae30 \uc704\ud55c \uc815\ucc45 \uc5d4\uc9c4\ub3c4 \uc124\uacc4\ub418\uc5c8\ub2e4. \ud558\uc9c0\ub9cc \uc774\uc885 \ub124\ud2b8\uc6cc\ud06c \ud658\uacbd\uc5d0\uc11c\ub294 \uc0ac\uc6a9\uc790\ub4e4\uc774 \uc694\uad6c\ud558\ub294 \uc11c\ub978 \ub2e4\ub978 \ub370\uc774\ud130 \uad6c\uc870\uc758 \uc131\ub2a5 \uc9c0\ud45c\uc640 \uc774\ub97c \ucc98\ub9ac\ud560 \uac01 \uc54c\uace0\ub9ac\uc998\uc758 \uc11c\ub85c \ub2e4\ub978 \uc2e4\ud589 \ud658\uacbd\uc5d0 \uc801\uc751\uc801\uc73c\ub85c \ub300\uc751\ud558\uae30 \uc5b4\ub824\uc6b4 \ubb38\uc81c\uac00 \uc788\uc5c8\ub2e4. \uc774\uc5d0 \ubcf8 \uc5f0\uad6c\uc5d0\uc11c\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4\uc640 \uc2e4\ud589 \ud658\uacbd \ub4f1 \ud2b9\uc815 \uae30\uc220\uc5d0 \uc885\uc18d\ub418\uc9c0 \uc54a\ub294 \uc815\ucc45 \uc5d4\uc9c4\uc744 \uc704\ud55c XML \uae30\ubc18\uc758 \uc778\ud130\ud398\uc774\uc2a4 \ud3ec\ub9f7\uc744 \uc815\uc758\ud558\uace0 \uadf8 \uc2a4\ud0a4\ub9c8\ub97c \uc81c\uc548\ud55c\ub2e4. \uc81c\uc548\ub41c \uc2a4\ud0a4\ub9c8\ub97c \uc0ac\uc6a9\ud558\uc5ec \uba54\uc2dc\uc9c0\ub294 \ud2b9\uc815 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4\uc5d0 \uc885\uc18d\ub418\uc9c0 \uc54a\uace0 \uc778\ucf54\ub529\uacfc \ub514\ucf54\ub529\uc744 \ud560 \uc218 \uc788\uc73c\uba70 Open Container Initiative \ud45c\uc900\uc744 \uae30\ubc18\uc73c\ub85c\uc2e4\ud589 \ud658\uacbd\uc744 \uc815\uc758\ud558\ub294 \ucee8\ud14c\uc774\ub108\ub97c \uae30\uc220\ud560 \uc218 \uc788\ub2e4.<br \/>\r\n<br \/>\r\nIn modern warfare environment, the well defined networks becomes important to the operations. Thus, researchers study on how to use the military tactical network resources effectively. To analyze effectiveness of the results from multiple studies together, an integrated testbed is critical as well as the design and the implementation of a policy engine that performs multiple network algorithms and analyze performance simultaneously. However, when the network environment is heterogeneous, it is hard to respond adaptively to the performance indicators of the different data structures and the different execution environments of each user algorithm. To address this issue, we propose an XML-based interface format and its schema for the policy engine, which is independent from specific technologies such as programming languages and execution environments. A message from and to the policy engine and the testbed can be encoded and decoded regardless of the programming language. Furthermore, it can describe containers of the execution environment based on the Open Container Initiative standard.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('160','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_160\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART002877194\" title=\"https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArti[...]\" target=\"_blank\">https:\/\/www.kci.go.kr\/kciportal\/ci\/sereArticleSearch\/ciSereArtiView.kci?sereArti[...]<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('160','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">12.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Park, Juwon;  Yoon, Daegun;  Yeo, Sangho;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('158','tp_links')\" style=\"cursor:pointer;\">AMBLE: Adjusting Mini-Batch and Local Epoch for Federated Learning with Heterogeneous Devices<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">Journal of Parallel and Distributed Computing, <\/span><span class=\"tp_pub_additional_year\">2022<\/span>, <span class=\"tp_pub_additional_issn\">ISSN: 0743-7315<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_158\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('158','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_158\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('158','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_158\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('158','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_158\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{Juwon2022AMBLE,<br \/>\r\ntitle = {AMBLE: Adjusting Mini-Batch and Local Epoch for Federated Learning with Heterogeneous Devices},<br \/>\r\nauthor = {Juwon Park and Daegun Yoon and Sangho Yeo and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0743731522001757},<br \/>\r\ndoi = {https:\/\/doi.org\/10.1016\/j.jpdc.2022.07.009},<br \/>\r\nissn = {0743-7315},<br \/>\r\nyear  = {2022},<br \/>\r\ndate = {2022-07-21},<br \/>\r\nurldate = {2022-07-21},<br \/>\r\njournal = {Journal of Parallel and Distributed Computing},<br \/>\r\nabstract = {As data privacy becomes increasingly important, federated learning applied to the training of deep learning models while ensuring the data privacy of devices is entering the spotlight. Federated learning makes it possible to process all data at once while processing data independently from various devices without collecting distributed local data in a central server. However, there are still challenges to overcome for the system of devices in federated learning such as communication overheads and the heterogeneity of the system. In this paper, we propose the Adjusting Mini-Batch and Local Epoch (AMBLE) approach, which adaptively adjusts the local mini-batch and local epoch size for heterogeneous devices in federated learning and updates the parameters synchronously. With AMBLE, we enhance the computational efficiency by removing stragglers and scaling the local learning rate to improve the model convergence rate and accuracy. We verify that federated learning with AMBLE is a stably trained model with a faster convergence speed and higher accuracy than FedAvg and adaptive batch size scheme for both identically and independently distributed (IID) and non-IID cases.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('158','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_158\" style=\"display:none;\"><div class=\"tp_abstract_entry\">As data privacy becomes increasingly important, federated learning applied to the training of deep learning models while ensuring the data privacy of devices is entering the spotlight. Federated learning makes it possible to process all data at once while processing data independently from various devices without collecting distributed local data in a central server. However, there are still challenges to overcome for the system of devices in federated learning such as communication overheads and the heterogeneity of the system. In this paper, we propose the Adjusting Mini-Batch and Local Epoch (AMBLE) approach, which adaptively adjusts the local mini-batch and local epoch size for heterogeneous devices in federated learning and updates the parameters synchronously. With AMBLE, we enhance the computational efficiency by removing stragglers and scaling the local learning rate to improve the model convergence rate and accuracy. We verify that federated learning with AMBLE is a stably trained model with a faster convergence speed and higher accuracy than FedAvg and adaptive batch size scheme for both identically and independently distributed (IID) and non-IID cases.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('158','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_158\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0743731522001757\" title=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0743731522001757\" target=\"_blank\">https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0743731522001757<\/a><\/li><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/https:\/\/doi.org\/10.1016\/j.jpdc.2022.07.009\" title=\"Follow DOI:https:\/\/doi.org\/10.1016\/j.jpdc.2022.07.009\" target=\"_blank\">doi:https:\/\/doi.org\/10.1016\/j.jpdc.2022.07.009<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('158','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">11.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('159','tp_links')\" style=\"cursor:pointer;\">SURF: Direction-Optimizing Breadth-First Search Using Workload State on GPUs<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">Sensors, <\/span><span class=\"tp_pub_additional_volume\">vol. 22, <\/span><span class=\"tp_pub_additional_number\">no. 13, <\/span><span class=\"tp_pub_additional_pages\">pp. 4899, <\/span><span class=\"tp_pub_additional_year\">2022<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_159\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('159','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_159\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('159','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_159\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('159','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_159\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{yoon2022surf,<br \/>\r\ntitle = {SURF: Direction-Optimizing Breadth-First Search Using Workload State on GPUs},<br \/>\r\nauthor = {Daegun Yoon and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/www.mdpi.com\/1424-8220\/22\/13\/4899},<br \/>\r\ndoi = {https:\/\/doi.org\/10.3390\/s22134899},<br \/>\r\nyear  = {2022},<br \/>\r\ndate = {2022-06-29},<br \/>\r\nurldate = {2022-01-01},<br \/>\r\njournal = {Sensors},<br \/>\r\nvolume = {22},<br \/>\r\nnumber = {13},<br \/>\r\npages = {4899},<br \/>\r\npublisher = {Multidisciplinary Digital Publishing Institute},<br \/>\r\nabstract = { Graph data structures have been used in a wide range of applications including scientific and social network applications. Engineers and scientists analyze graph data to discover knowledge and insights by using various graph algorithms. A breadth-first search (BFS) is one of the fundamental building blocks of complex graph algorithms and its implementation is included in graph libraries for large-scale graph processing. In this paper, we propose a novel direction selection method, SURF (Selecting directions Upon Recent workload of Frontiers) to enhance the performance of BFS on GPU. A direction optimization that selects the proper traversal direction of a BFS execution between the push and pull phases is crucial to the performance as well as for efficient handling of the varying workloads of the frontiers. However, existing works select the direction using condition statements based on predefined thresholds without considering the changing workload state. To solve this drawback, we define several metrics that describe the state of the workload and analyze their impact on the BFS performance. To show that SURF selects the appropriate direction, we implement the direction selection method with a deep neural network model that adopts those metrics as the input features. Experimental results indicate that SURF achieves a higher direction prediction accuracy and reduced execution time in comparison with existing state-of-the-art methods that support a direction-optimizing BFS. SURF yields up to a 5.62\u00d7},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('159','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_159\" style=\"display:none;\"><div class=\"tp_abstract_entry\"> Graph data structures have been used in a wide range of applications including scientific and social network applications. Engineers and scientists analyze graph data to discover knowledge and insights by using various graph algorithms. A breadth-first search (BFS) is one of the fundamental building blocks of complex graph algorithms and its implementation is included in graph libraries for large-scale graph processing. In this paper, we propose a novel direction selection method, SURF (Selecting directions Upon Recent workload of Frontiers) to enhance the performance of BFS on GPU. A direction optimization that selects the proper traversal direction of a BFS execution between the push and pull phases is crucial to the performance as well as for efficient handling of the varying workloads of the frontiers. However, existing works select the direction using condition statements based on predefined thresholds without considering the changing workload state. To solve this drawback, we define several metrics that describe the state of the workload and analyze their impact on the BFS performance. To show that SURF selects the appropriate direction, we implement the direction selection method with a deep neural network model that adopts those metrics as the input features. Experimental results indicate that SURF achieves a higher direction prediction accuracy and reduced execution time in comparison with existing state-of-the-art methods that support a direction-optimizing BFS. SURF yields up to a 5.62\u00d7<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('159','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_159\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.mdpi.com\/1424-8220\/22\/13\/4899\" title=\"https:\/\/www.mdpi.com\/1424-8220\/22\/13\/4899\" target=\"_blank\">https:\/\/www.mdpi.com\/1424-8220\/22\/13\/4899<\/a><\/li><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/https:\/\/doi.org\/10.3390\/s22134899\" title=\"Follow DOI:https:\/\/doi.org\/10.3390\/s22134899\" target=\"_blank\">doi:https:\/\/doi.org\/10.3390\/s22134899<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('159','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">10.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \uc815\ud604\uc11d,;  \uc720\ubbf8\ub9ac,;  \uc724\ub300\uac74,;  \uc774\uc2b9\uc900,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('156','tp_links')\" style=\"cursor:pointer;\">\uc7ac\ub09c \ub300\uc751 \uae30\uacc4\ud559\uc2b5 \ubaa8\ub378\uc758 Data Drift \ubb38\uc81c\uc5d0 \ub300\ud55c MLOps \uae30\ubc18 \ub300\uc751 \uae30\ubc95<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">2022 \ud55c\uad6d\ucc28\uc138\ub300\ucef4\ud4e8\ud305\ud559\ud68c \ucd98\uacc4\ud559\uc220\ub300\ud68c, <\/span><span class=\"tp_pub_additional_publisher\">\ud55c\uad6d\ucc28\uc138\ub300\ucef4\ud4e8\ud305\ud559\ud68c, <\/span><span class=\"tp_pub_additional_year\">2022<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_156\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('156','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_156\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('156','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_156\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('156','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_156\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{\uc815\ud604\uc11d2022\uc7ac\ub09c,<br \/>\r\ntitle = {\uc7ac\ub09c \ub300\uc751 \uae30\uacc4\ud559\uc2b5 \ubaa8\ub378\uc758 Data Drift \ubb38\uc81c\uc5d0 \ub300\ud55c MLOps \uae30\ubc18 \ub300\uc751 \uae30\ubc95},<br \/>\r\nauthor = {\uc815\ud604\uc11d and \uc720\ubbf8\ub9ac and \uc724\ub300\uac74 and \uc774\uc2b9\uc900 and \uc624\uc0c1\uc724},<br \/>\r\nurl = {https:\/\/www.earticle.net\/Article\/A412404},<br \/>\r\nyear  = {2022},<br \/>\r\ndate = {2022-05-21},<br \/>\r\nurldate = {2022-05-21},<br \/>\r\nbooktitle = {2022 \ud55c\uad6d\ucc28\uc138\ub300\ucef4\ud4e8\ud305\ud559\ud68c \ucd98\uacc4\ud559\uc220\ub300\ud68c},<br \/>\r\npages = {pp.473-476},<br \/>\r\npublisher = {\ud55c\uad6d\ucc28\uc138\ub300\ucef4\ud4e8\ud305\ud559\ud68c},<br \/>\r\nabstract = {\uae30\uacc4\ud559\uc2b5\uc5d0\uc11c Data Drift\ub294 \uc815\ud655\ub3c4\uc5d0 \ud070 \uc601\ud5a5\uc744 \uc8fc\ub294 \uc911\uc694\ud55c \ubb38\uc81c\uc774\uba70, \uc7ac\ub09c \ub300\uc751\uacfc \uac19\uc774 \ubaa8\ub378\uc758 \uc798\ubabb\ub41c \uc608\uce21 \ud53c\ud574\uac00 \ud070 \ubd84\uc57c\uc5d0\uc11c \ub354 \uc911\uc694\ud558\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294 \uc7ac\ub09c \ubd84\uc57c Data Drift \ubb38\uc81c\uc5d0 \ub300\ud574 MLOps\ub97c \uc774\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \uc7ac\ud559\uc2b5\uc744 \ud6a8\uacfc\uc801\uc73c\ub85c \uc218\ud589\ud560 \uc218 \uc788\ub294 \ubc29\uc548\uc73c\ub85c MLOps \uae30\ubc95\uacfc \ud234\ub4e4\uc744 \uc0ac\uc6a9\ud558\ub294 \uac83\uc744 \uc81c\uc548\ud558\uace0, Kaggle \ub370\uc774\ud130\uc640 MLFlow\ub97c \uae30\ubc18\uc73c\ub85c \uc815\ud655\ub3c4 \uc2e4\ud5d8\uc744 \uc218\ud589\ud558\uc5ec \uc8fc\uc7a5\uc744 \uac80\uc99d\ud558\uc600\ub2e4.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('156','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_156\" style=\"display:none;\"><div class=\"tp_abstract_entry\">\uae30\uacc4\ud559\uc2b5\uc5d0\uc11c Data Drift\ub294 \uc815\ud655\ub3c4\uc5d0 \ud070 \uc601\ud5a5\uc744 \uc8fc\ub294 \uc911\uc694\ud55c \ubb38\uc81c\uc774\uba70, \uc7ac\ub09c \ub300\uc751\uacfc \uac19\uc774 \ubaa8\ub378\uc758 \uc798\ubabb\ub41c \uc608\uce21 \ud53c\ud574\uac00 \ud070 \ubd84\uc57c\uc5d0\uc11c \ub354 \uc911\uc694\ud558\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294 \uc7ac\ub09c \ubd84\uc57c Data Drift \ubb38\uc81c\uc5d0 \ub300\ud574 MLOps\ub97c \uc774\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \uc7ac\ud559\uc2b5\uc744 \ud6a8\uacfc\uc801\uc73c\ub85c \uc218\ud589\ud560 \uc218 \uc788\ub294 \ubc29\uc548\uc73c\ub85c MLOps \uae30\ubc95\uacfc \ud234\ub4e4\uc744 \uc0ac\uc6a9\ud558\ub294 \uac83\uc744 \uc81c\uc548\ud558\uace0, Kaggle \ub370\uc774\ud130\uc640 MLFlow\ub97c \uae30\ubc18\uc73c\ub85c \uc815\ud655\ub3c4 \uc2e4\ud5d8\uc744 \uc218\ud589\ud558\uc5ec \uc8fc\uc7a5\uc744 \uac80\uc99d\ud558\uc600\ub2e4.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('156','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_156\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.earticle.net\/Article\/A412404\" title=\"https:\/\/www.earticle.net\/Article\/A412404\" target=\"_blank\">https:\/\/www.earticle.net\/Article\/A412404<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('156','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><h3 class=\"tp_h3\" id=\"tp_h3_2021\">2021<\/h3><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">9.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Lee, Seungjun;  Yoon, Daegun;  Yeo, Sangho;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('155','tp_links')\" style=\"cursor:pointer;\">Mitigating Cold Start Problem in Serverless Computing with Function Fusion<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">Sensors, <\/span><span class=\"tp_pub_additional_volume\">vol. 21, <\/span><span class=\"tp_pub_additional_number\">no. 24, <\/span><span class=\"tp_pub_additional_year\">2021<\/span>, <span class=\"tp_pub_additional_issn\">ISSN: 1424-8220<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_155\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('155','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_155\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('155','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_155\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('155','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_155\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{s21248416,<br \/>\r\ntitle = {Mitigating Cold Start Problem in Serverless Computing with Function Fusion},<br \/>\r\nauthor = {Seungjun Lee and Daegun Yoon and Sangho Yeo and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/www.mdpi.com\/1424-8220\/21\/24\/8416},<br \/>\r\ndoi = {10.3390\/s21248416},<br \/>\r\nissn = {1424-8220},<br \/>\r\nyear  = {2021},<br \/>\r\ndate = {2021-12-23},<br \/>\r\nurldate = {2021-12-16},<br \/>\r\njournal = {Sensors},<br \/>\r\nvolume = {21},<br \/>\r\nnumber = {24},<br \/>\r\nabstract = {As Artificial Intelligence (AI) is becoming ubiquitous in many applications, serverless computing is also emerging as a building block for developing cloud-based AI services. Serverless computing has received much interest because of its simplicity, scalability, and resource efficiency. However, due to the trade-off with resource efficiency, serverless computing suffers from the cold start problem, that is, a latency between a request arrival and function execution. The cold start problem significantly influences the overall response time of workflow that consists of functions because the cold start may occur in every function within the workflow. Function fusion can be one of the solutions to mitigate the cold start latency of a workflow. If two functions are fused into a single function, the cold start of the second function is removed; however, if parallel functions are fused, the workflow response time can be increased because the parallel functions run sequentially even if the cold start latency is reduced. This study presents an approach to mitigate the cold start latency of a workflow using function fusion while considering a parallel run. First, we identify three latencies that affect response time, present a workflow response time model considering the latency, and efficiently find a fusion solution that can optimize the response time on the cold start. Our method shows a response time of 28%\u201386% of the response time of the original workflow in five workflows.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('155','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_155\" style=\"display:none;\"><div class=\"tp_abstract_entry\">As Artificial Intelligence (AI) is becoming ubiquitous in many applications, serverless computing is also emerging as a building block for developing cloud-based AI services. Serverless computing has received much interest because of its simplicity, scalability, and resource efficiency. However, due to the trade-off with resource efficiency, serverless computing suffers from the cold start problem, that is, a latency between a request arrival and function execution. The cold start problem significantly influences the overall response time of workflow that consists of functions because the cold start may occur in every function within the workflow. Function fusion can be one of the solutions to mitigate the cold start latency of a workflow. If two functions are fused into a single function, the cold start of the second function is removed; however, if parallel functions are fused, the workflow response time can be increased because the parallel functions run sequentially even if the cold start latency is reduced. This study presents an approach to mitigate the cold start latency of a workflow using function fusion while considering a parallel run. First, we identify three latencies that affect response time, present a workflow response time model considering the latency, and efficiently find a fusion solution that can optimize the response time on the cold start. Our method shows a response time of 28%\u201386% of the response time of the original workflow in five workflows.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('155','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_155\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.mdpi.com\/1424-8220\/21\/24\/8416\" title=\"https:\/\/www.mdpi.com\/1424-8220\/21\/24\/8416\" target=\"_blank\">https:\/\/www.mdpi.com\/1424-8220\/21\/24\/8416<\/a><\/li><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/10.3390\/s21248416\" title=\"Follow DOI:10.3390\/s21248416\" target=\"_blank\">doi:10.3390\/s21248416<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('155','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">8.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('149','tp_links')\" style=\"cursor:pointer;\">Traversing Large Road Networks on GPUs with Breadth-First Search<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">The 7th International Conference on Next Generation Computing, <\/span><span class=\"tp_pub_additional_year\">2021<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_149\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('149','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_149\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('149','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_149\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('149','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_149\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{Yoon2021Traversing,<br \/>\r\ntitle = {Traversing Large Road Networks on GPUs with Breadth-First Search},<br \/>\r\nauthor = {Daegun Yoon and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/www.earticle.net\/Article\/A448039},<br \/>\r\nyear  = {2021},<br \/>\r\ndate = {2021-11-05},<br \/>\r\nurldate = {2021-11-05},<br \/>\r\nbooktitle = {The 7th International Conference on Next Generation Computing},<br \/>\r\njournal = {The 7th International Conference on Next Generation Computing 2021},<br \/>\r\nabstract = {Breadth-first search (BFS) is one of the most used graph kernels, and substantially affects the overall performance when processing various graphs. Since graph data are frequently used in real life for example road networks in navigation systems, high performance graph processing becomes more critical. In this study, we aim to process BFS algorithm efficiently on road network data. We propose BARON, a BFS algorithm that copes with road networks. To accelerate graph traversal, BARON reduce the occurrence of branch and memory divergences by exploiting warp-cooperative work sharing and atomic operations. With this design approach, BARON outperforms the other BFS kernels of state-of-the-art graph processing frameworks executed stably on the latest GPU architectures. For various graphs, BARON yields speedups of up to 2.88x and 5.43x over Gunrock and CuSha, respectively.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('149','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_149\" style=\"display:none;\"><div class=\"tp_abstract_entry\">Breadth-first search (BFS) is one of the most used graph kernels, and substantially affects the overall performance when processing various graphs. Since graph data are frequently used in real life for example road networks in navigation systems, high performance graph processing becomes more critical. In this study, we aim to process BFS algorithm efficiently on road network data. We propose BARON, a BFS algorithm that copes with road networks. To accelerate graph traversal, BARON reduce the occurrence of branch and memory divergences by exploiting warp-cooperative work sharing and atomic operations. With this design approach, BARON outperforms the other BFS kernels of state-of-the-art graph processing frameworks executed stably on the latest GPU architectures. For various graphs, BARON yields speedups of up to 2.88x and 5.43x over Gunrock and CuSha, respectively.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('149','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_149\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.earticle.net\/Article\/A448039\" title=\"https:\/\/www.earticle.net\/Article\/A448039\" target=\"_blank\">https:\/\/www.earticle.net\/Article\/A448039<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('149','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">7.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Lee, Seungjun;  Yoon, Daegun;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('146','tp_links')\" style=\"cursor:pointer;\">Imitation learning for VM placement problem using demonstration data generated by heuristics<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">17th Int. Conference on Data Science (ICDATA\u201921), <\/span><span class=\"tp_pub_additional_year\">2021<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_146\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('146','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_146\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('146','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_146\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('146','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_146\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{lee2021imitation,<br \/>\r\ntitle = {Imitation learning for VM placement problem using demonstration data generated by heuristics},<br \/>\r\nauthor = {Seungjun Lee and Daegun Yoon and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/youtu.be\/CmG3E1rWroQ},<br \/>\r\nyear  = {2021},<br \/>\r\ndate = {2021-07-26},<br \/>\r\nurldate = {2021-07-26},<br \/>\r\nbooktitle = {17th Int. Conference on Data Science (ICDATA\u201921)},<br \/>\r\nabstract = {Data centers are key components of cloud computing to run virtual machines. For saving the cost to operate data centers, it is important to decide how to allocate each virtual machine to a certain physical machine. Because the virtual machine placement problem is NP-Hard, there are many heuristics to obtain near-optimal solutions as quickly as possible. The reinforcement learning technique can be applied for virtual machine placement problem. However, if the problem size gets bigger, the convergence speed of reinforcement learning gets slower. The possible solution is that the agent imitates the behavior of given demonstration, called imitation learning. In this paper, we propose a method combining reinforcement learning with imitation learning. In our proposed approach, demonstration data is generated by simple heuristics not human experts.},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('146','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_146\" style=\"display:none;\"><div class=\"tp_abstract_entry\">Data centers are key components of cloud computing to run virtual machines. For saving the cost to operate data centers, it is important to decide how to allocate each virtual machine to a certain physical machine. Because the virtual machine placement problem is NP-Hard, there are many heuristics to obtain near-optimal solutions as quickly as possible. The reinforcement learning technique can be applied for virtual machine placement problem. However, if the problem size gets bigger, the convergence speed of reinforcement learning gets slower. The possible solution is that the agent imitates the behavior of given demonstration, called imitation learning. In this paper, we propose a method combining reinforcement learning with imitation learning. In our proposed approach, demonstration data is generated by simple heuristics not human experts.<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('146','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_146\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fab fa-youtube\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/youtu.be\/CmG3E1rWroQ\" title=\"https:\/\/youtu.be\/CmG3E1rWroQ\" target=\"_blank\">https:\/\/youtu.be\/CmG3E1rWroQ<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('146','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">6.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \uc724\ub300\uac74,;  \ub178\ubcd1\ud76c,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\">\uc804\uc220\ub9dd  \uc131\ub2a5  \uac1c\ub7c9\uc744  \uc704\ud55c  \uc815\ucc45  \uc5d4\uc9c4  \uc778\ud130\ud398\uc774\uc2a4  \uc124\uacc4 <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">2021 \ud55c\uad6d\uad70\uc0ac\uacfc\ud559\uae30\uc220\ud559\ud68c \uc885\ud569\ud559\uc220\ub300\ud68c, <\/span><span class=\"tp_pub_additional_year\">2021<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_148\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('148','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_148\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{\uc724\ub300\uac742021\uc804\uc220\ub9dd,<br \/>\r\ntitle = {\uc804\uc220\ub9dd  \uc131\ub2a5  \uac1c\ub7c9\uc744  \uc704\ud55c  \uc815\ucc45  \uc5d4\uc9c4  \uc778\ud130\ud398\uc774\uc2a4  \uc124\uacc4},<br \/>\r\nauthor = {\uc724\ub300\uac74 and \ub178\ubcd1\ud76c and \uc624\uc0c1\uc724},<br \/>\r\nyear  = {2021},<br \/>\r\ndate = {2021-06-10},<br \/>\r\nbooktitle = {2021 \ud55c\uad6d\uad70\uc0ac\uacfc\ud559\uae30\uc220\ud559\ud68c \uc885\ud569\ud559\uc220\ub300\ud68c},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('148','tp_bibtex')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">5.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Li, Zhetao;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('141','tp_links')\" style=\"cursor:pointer;\">Balanced content space partitioning for pub\/sub: a study on impact of varying partitioning granularity<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">The Journal of Supercomputing, <\/span><span class=\"tp_pub_additional_pages\">pp. 1\u201327, <\/span><span class=\"tp_pub_additional_year\">2021<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_resource_link\"><a id=\"tp_links_sh_141\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('141','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_141\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('141','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_141\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{yoon2021balanced,<br \/>\r\ntitle = {Balanced content space partitioning for pub\/sub: a study on impact of varying partitioning granularity},<br \/>\r\nauthor = {Daegun Yoon and Zhetao Li and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/link.springer.com\/article\/10.1007\/s11227-021-03821-5},<br \/>\r\nyear  = {2021},<br \/>\r\ndate = {2021-01-01},<br \/>\r\njournal = {The Journal of Supercomputing},<br \/>\r\npages = {1--27},<br \/>\r\npublisher = {Springer},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('141','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_141\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/link.springer.com\/article\/10.1007\/s11227-021-03821-5\" title=\"https:\/\/link.springer.com\/article\/10.1007\/s11227-021-03821-5\" target=\"_blank\">https:\/\/link.springer.com\/article\/10.1007\/s11227-021-03821-5<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('141','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><h3 class=\"tp_h3\" id=\"tp_h3_2020\">2020<\/h3><div class=\"tp_publication tp_publication_article\"><div class=\"tp_pub_number\">4.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Park, Gyudong;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('1','tp_links')\" style=\"cursor:pointer;\">Exploring a system architecture of content-based publish\/subscribe system for efficient on-the-fly data dissemination<\/a> <span class=\"tp_pub_type tp_  article\">Journal Article<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_in\">In: <\/span><span class=\"tp_pub_additional_journal\">Concurrency and Computation: Practice and Experience, <\/span><span class=\"tp_pub_additional_pages\">pp. e6090, <\/span><span class=\"tp_pub_additional_year\">2020<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_resource_link\"><a id=\"tp_links_sh_1\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('1','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_1\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('1','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_1\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@article{yoon2020exploring,<br \/>\r\ntitle = {Exploring a system architecture of content-based publish\/subscribe system for efficient on-the-fly data dissemination},<br \/>\r\nauthor = {Daegun Yoon and Gyudong Park and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/onlinelibrary.wiley.com\/doi\/full\/10.1002\/cpe.6090},<br \/>\r\nyear  = {2020},<br \/>\r\ndate = {2020-01-01},<br \/>\r\njournal = {Concurrency and Computation: Practice and Experience},<br \/>\r\npages = {e6090},<br \/>\r\npublisher = {Wiley Online Library},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {article}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('1','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_1\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/onlinelibrary.wiley.com\/doi\/full\/10.1002\/cpe.6090\" title=\"https:\/\/onlinelibrary.wiley.com\/doi\/full\/10.1002\/cpe.6090\" target=\"_blank\">https:\/\/onlinelibrary.wiley.com\/doi\/full\/10.1002\/cpe.6090<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('1','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">3.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> Yoon, Daegun;  Park, Gyudong;  Oh, Sangyoon<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('11','tp_links')\" style=\"cursor:pointer;\">CPartition: a Correlation-Based Space Partitioning for Content-Based Publish\/Subscribe Systems with Skewed Workload<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">2020 IEEE International Conference on Big Data and Smart Computing (BigComp), <\/span><span class=\"tp_pub_additional_organization\">IEEE <\/span><span class=\"tp_pub_additional_year\">2020<\/span>, <span class=\"tp_pub_additional_isbn\">ISBN: 978-1-7281-6034-4<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_resource_link\"><a id=\"tp_links_sh_11\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('11','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_11\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('11','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_11\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{yoon2020cpartition,<br \/>\r\ntitle = {CPartition: a Correlation-Based Space Partitioning for Content-Based Publish\/Subscribe Systems with Skewed Workload},<br \/>\r\nauthor = {Daegun Yoon and Gyudong Park and Sangyoon Oh},<br \/>\r\nurl = {https:\/\/ieeexplore.ieee.org\/abstract\/document\/9070640},<br \/>\r\ndoi = {10.1109\/BigComp48618.2020.00-46},<br \/>\r\nisbn = {978-1-7281-6034-4},<br \/>\r\nyear  = {2020},<br \/>\r\ndate = {2020-01-01},<br \/>\r\nurldate = {2020-01-01},<br \/>\r\nbooktitle = {2020 IEEE International Conference on Big Data and Smart Computing (BigComp)},<br \/>\r\npages = {377--384},<br \/>\r\norganization = {IEEE},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('11','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_11\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/ieeexplore.ieee.org\/abstract\/document\/9070640\" title=\"https:\/\/ieeexplore.ieee.org\/abstract\/document\/9070640\" target=\"_blank\">https:\/\/ieeexplore.ieee.org\/abstract\/document\/9070640<\/a><\/li><li><i class=\"ai ai-doi\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/dx.doi.org\/10.1109\/BigComp48618.2020.00-46\" title=\"Follow DOI:10.1109\/BigComp48618.2020.00-46\" target=\"_blank\">doi:10.1109\/BigComp48618.2020.00-46<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('11','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">2.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \uc724\ub300\uac74,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\"><a class=\"tp_title_link\" onclick=\"teachpress_pub_showhide('2','tp_links')\" style=\"cursor:pointer;\">Software-Defined Network \uc5d0\uc11c\uc758 Conflict Resolution \uc744 \uc704\ud55c \uc815\ucc45\uc5d4\uc9c4 \uad6c\uc870 \ubc0f \uc804\ub7b5 \ubd84\uc11d<\/a> <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ud559\uc220\ub300\ud68c\ub17c\ubb38\uc9d1, <\/span><span class=\"tp_pub_additional_year\">2020<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_abstract_link\"><a id=\"tp_abstract_sh_2\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('2','tp_abstract')\" title=\"Show abstract\" style=\"cursor:pointer;\">Abstract<\/a><\/span> | <span class=\"tp_resource_link\"><a id=\"tp_links_sh_2\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('2','tp_links')\" title=\"Show links and resources\" style=\"cursor:pointer;\">Links<\/a><\/span> | <span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_2\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('2','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_2\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{\uc724\ub300\uac742020software,<br \/>\r\ntitle = {Software-Defined Network \uc5d0\uc11c\uc758 Conflict Resolution \uc744 \uc704\ud55c \uc815\ucc45\uc5d4\uc9c4 \uad6c\uc870 \ubc0f \uc804\ub7b5 \ubd84\uc11d},<br \/>\r\nauthor = {\uc724\ub300\uac74 and \uc624\uc0c1\uc724},<br \/>\r\nurl = {https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE10498659&mark=0&useDate=&ipRange=N&accessgl=Y&language=ko_KR&hasTopBanner=false},<br \/>\r\nyear  = {2020},<br \/>\r\ndate = {2020-01-01},<br \/>\r\nurldate = {2020-01-01},<br \/>\r\nbooktitle = {\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ud559\uc220\ub300\ud68c\ub17c\ubb38\uc9d1},<br \/>\r\njournal = {\ud55c\uad6d\ud1b5\uc2e0\ud559\ud68c \ud559\uc220\ub300\ud68c\ub17c\ubb38\uc9d1},<br \/>\r\npages = {566--567},<br \/>\r\nabstract = {\ubcf8  \ub17c\ubb38\uc740  Software-Defined  Network \uc5d0\uc11c  \ucd94\uac00\ub418\ub294  \uc815\ucc45\ub4e4\uc774  \uae30\uc874\uc758  \uc815\ucc45\ub4e4\uacfc  \uc77c\uc73c\ud0a4\ub294  \ucda9\ub3cc\uc744  \ud574\uacb0\ud558\uae30 \uc704\ud574  \uc874\uc7ac\ud558\ub294  \uc815\ucc45\uc5d4\uc9c4\uc758  \uc0c8\ub85c\uc6b4  \uc124\uacc4\ub97c  \uc704\ud574  \ubd84\uc11d\ud55c  \uae30\uc874  \uc5f0\uad6c  \ub0b4\uc6a9\ub4e4\uc744  \uc18c\uac1c\ud558\uace0  \ubd84\uc11d\ud55c  \ub0b4\uc6a9\ub4e4\uc744 \uae30\ubc18\uc73c\ub85c  \uc815\ucc45\uc5d4\uc9c4\uc758  conflict  resolution  \uc804\ub7b5\uc744  \uc81c\uc548\ud55c\ub2e4.  \ubcf8  \ub17c\ubb38\uc5d0\uc11c\ub294  SDN \uc5d0  \uc0c8\ub85c\uc6b4  \uc815\ucc45\uc774  \ucd94\uac00\ub418\ub294 \uacbd\uc6b0  \ubc1c\uc0dd\ud560  \uc218  \uc788\ub294  conflict \ub97c  \uac10\uc9c0\ud55c  \ud6c4  \ud574\uacb0\ud558\uae30  \uc704\ud574,  \uc815\ucc45\uc5d4\uc9c4\uc774  conflict  detector \uc640  conflict handler \ub85c  \uad6c\uc131\ub418\ub294  \uad6c\uc870\ub97c  \uac00\uc815\ud55c\ub2e4.  Conflict  detector \ub294  \uc0c8\ub85c  \ucd94\uac00\ub418\ub294  \uc815\ucc45\uc774  \uae30\uc874\uc758  \uc815\ucc45\ub4e4\uacfc  \ucda9\ub3cc\uc744 \uc77c\uc73c\ud0a4\ub294\uc9c0  \uac10\uc9c0\ud558\uace0  conflict  handler \ub294  conflict  resolution \uc744  \ud1b5\ud574  \ubb38\uc81c\uac00  \ub418\ub294  \uc815\ucc45\uc744  \uc0ad\uc81c\ud558\ub294  \uc5ed\ud560\uc744 \ud55c\ub2e4.  \ubcf8  \ub17c\ubb38\uc5d0\uc11c\ub294  conflict  handler \uac00  Recency  \uc911\uc2ec  \uc804\ub7b5\uacfc  Priority  \uc911\uc2ec  \uc804\ub7b5\uc744  \uc0ac\uc6a9\ud558\uc5ec  \ubb38\uc81c\uac00  \ub418\ub294 \uc815\ucc45\uc744  \uc0ad\uc81c\ud558\ub294  \ubc29\uc548\uc5d0  \ub300\ud574\uc11c  \ubd84\uc11d\ud55c  \uacb0\uacfc\ub97c  \uc18c\uac1c\ud55c\ub2e4},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('2','tp_bibtex')\">Close<\/a><\/p><\/div><div class=\"tp_abstract\" id=\"tp_abstract_2\" style=\"display:none;\"><div class=\"tp_abstract_entry\">\ubcf8  \ub17c\ubb38\uc740  Software-Defined  Network \uc5d0\uc11c  \ucd94\uac00\ub418\ub294  \uc815\ucc45\ub4e4\uc774  \uae30\uc874\uc758  \uc815\ucc45\ub4e4\uacfc  \uc77c\uc73c\ud0a4\ub294  \ucda9\ub3cc\uc744  \ud574\uacb0\ud558\uae30 \uc704\ud574  \uc874\uc7ac\ud558\ub294  \uc815\ucc45\uc5d4\uc9c4\uc758  \uc0c8\ub85c\uc6b4  \uc124\uacc4\ub97c  \uc704\ud574  \ubd84\uc11d\ud55c  \uae30\uc874  \uc5f0\uad6c  \ub0b4\uc6a9\ub4e4\uc744  \uc18c\uac1c\ud558\uace0  \ubd84\uc11d\ud55c  \ub0b4\uc6a9\ub4e4\uc744 \uae30\ubc18\uc73c\ub85c  \uc815\ucc45\uc5d4\uc9c4\uc758  conflict  resolution  \uc804\ub7b5\uc744  \uc81c\uc548\ud55c\ub2e4.  \ubcf8  \ub17c\ubb38\uc5d0\uc11c\ub294  SDN \uc5d0  \uc0c8\ub85c\uc6b4  \uc815\ucc45\uc774  \ucd94\uac00\ub418\ub294 \uacbd\uc6b0  \ubc1c\uc0dd\ud560  \uc218  \uc788\ub294  conflict \ub97c  \uac10\uc9c0\ud55c  \ud6c4  \ud574\uacb0\ud558\uae30  \uc704\ud574,  \uc815\ucc45\uc5d4\uc9c4\uc774  conflict  detector \uc640  conflict handler \ub85c  \uad6c\uc131\ub418\ub294  \uad6c\uc870\ub97c  \uac00\uc815\ud55c\ub2e4.  Conflict  detector \ub294  \uc0c8\ub85c  \ucd94\uac00\ub418\ub294  \uc815\ucc45\uc774  \uae30\uc874\uc758  \uc815\ucc45\ub4e4\uacfc  \ucda9\ub3cc\uc744 \uc77c\uc73c\ud0a4\ub294\uc9c0  \uac10\uc9c0\ud558\uace0  conflict  handler \ub294  conflict  resolution \uc744  \ud1b5\ud574  \ubb38\uc81c\uac00  \ub418\ub294  \uc815\ucc45\uc744  \uc0ad\uc81c\ud558\ub294  \uc5ed\ud560\uc744 \ud55c\ub2e4.  \ubcf8  \ub17c\ubb38\uc5d0\uc11c\ub294  conflict  handler \uac00  Recency  \uc911\uc2ec  \uc804\ub7b5\uacfc  Priority  \uc911\uc2ec  \uc804\ub7b5\uc744  \uc0ac\uc6a9\ud558\uc5ec  \ubb38\uc81c\uac00  \ub418\ub294 \uc815\ucc45\uc744  \uc0ad\uc81c\ud558\ub294  \ubc29\uc548\uc5d0  \ub300\ud574\uc11c  \ubd84\uc11d\ud55c  \uacb0\uacfc\ub97c  \uc18c\uac1c\ud55c\ub2e4<\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('2','tp_abstract')\">Close<\/a><\/p><\/div><div class=\"tp_links\" id=\"tp_links_2\" style=\"display:none;\"><div class=\"tp_links_entry\"><ul class=\"tp_pub_list\"><li><i class=\"fas fa-globe\"><\/i><a class=\"tp_pub_list\" href=\"https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE10498659&amp;mark=0&amp;useDate=&amp;ipRange=N&amp;accessgl=Y&amp;language=ko_KR&amp;hasTopBanner=false\" title=\"https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE10498659&amp;mark=0&amp;useDat[...]\" target=\"_blank\">https:\/\/www.dbpia.co.kr\/pdf\/pdfView.do?nodeId=NODE10498659&amp;mark=0&amp;useDat[...]<\/a><\/li><\/ul><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('2','tp_links')\">Close<\/a><\/p><\/div><\/div><\/div><h3 class=\"tp_h3\" id=\"tp_h3_2018\">2018<\/h3><div class=\"tp_publication tp_publication_conference\"><div class=\"tp_pub_number\">1.<\/div><div class=\"tp_pub_info\"><p class=\"tp_pub_author\"> \uc724\ub300\uac74,;  \uc624\uc0c1\uc724,<\/p><p class=\"tp_pub_title\">\ub3d9\uc885 \uc6b4\uc601\uccb4\uc81c \ud658\uacbd\uc5d0\uc11c\uc758 \uac00\uc0c1 \uba38\uc2e0 \ub9c8\uc774\uadf8\ub808\uc774\uc158 \uc131\ub2a5 \ubd84\uc11d <span class=\"tp_pub_type tp_  conference\">Conference<\/span> <\/p><p class=\"tp_pub_additional\"><span class=\"tp_pub_additional_booktitle\">\ud55c\uad6d\uc815\ubcf4\uacfc\ud559\ud68c KCC \ud559\uc220\ub300\ud68c, <\/span><span class=\"tp_pub_additional_year\">2018<\/span>.<\/p><p class=\"tp_pub_menu\"><span class=\"tp_bibtex_link\"><a id=\"tp_bibtex_sh_27\" class=\"tp_show\" onclick=\"teachpress_pub_showhide('27','tp_bibtex')\" title=\"Show BibTeX entry\" style=\"cursor:pointer;\">BibTeX<\/a><\/span><\/p><div class=\"tp_bibtex\" id=\"tp_bibtex_27\" style=\"display:none;\"><div class=\"tp_bibtex_entry\"><pre>@conference{\uc724\ub300\uac742018\ub3d9\uc885,<br \/>\r\ntitle = {\ub3d9\uc885 \uc6b4\uc601\uccb4\uc81c \ud658\uacbd\uc5d0\uc11c\uc758 \uac00\uc0c1 \uba38\uc2e0 \ub9c8\uc774\uadf8\ub808\uc774\uc158 \uc131\ub2a5 \ubd84\uc11d},<br \/>\r\nauthor = {\uc724\ub300\uac74 and \uc624\uc0c1\uc724},<br \/>\r\nyear  = {2018},<br \/>\r\ndate = {2018-01-01},<br \/>\r\nurldate = {2018-01-01},<br \/>\r\nbooktitle = {\ud55c\uad6d\uc815\ubcf4\uacfc\ud559\ud68c KCC \ud559\uc220\ub300\ud68c},<br \/>\r\npages = {47--49},<br \/>\r\nkeywords = {},<br \/>\r\npubstate = {published},<br \/>\r\ntppubtype = {conference}<br \/>\r\n}<br \/>\r\n<\/pre><\/div><p class=\"tp_close_menu\"><a class=\"tp_close\" onclick=\"teachpress_pub_showhide('27','tp_bibtex')\">Close<\/a><\/p><\/div><\/div><\/div><\/div><\/div>\n","protected":false},"excerpt":{"rendered":"<p>\uc724\ub300\uac74Daegun Yoon Email kljp at ajou.ac.kr Research interests Machine Learning, Parallel Algorithm, Distributed System Introduction Daegun Yoon is a Ph.D. candidate at the Department of Artificial Intelligence, Ajou University. He received the B.S. in the Department of Software, Ajou University, in 2018. Publications<\/p>\n","protected":false},"author":1,"featured_media":0,"parent":785,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_uag_custom_page_level_css":"","footnotes":""},"class_list":["post-851","page","type-page","status-publish","hentry"],"uagb_featured_image_src":{"full":false,"thumbnail":false,"medium":false,"medium_large":false,"large":false,"1536x1536":false,"2048x2048":false,"twentyseventeen-featured-image":false,"twentyseventeen-thumbnail-avatar":false},"uagb_author_info":{"display_name":"wise","author_link":"https:\/\/wise.ajou.ac.kr\/?author=1"},"uagb_comment_info":0,"uagb_excerpt":"\uc724\ub300\uac74Daegun Yoon Email kljp at ajou.ac.kr Research interests Machine Learning, Parallel Algorithm, Distributed System Introduction Daegun Yoon is a Ph.D. candidate at the Department of Artificial Intelligence, Ajou University. He received the B.S. in the Department of Software, Ajou University, in 2018. Publications","_links":{"self":[{"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=\/wp\/v2\/pages\/851","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=851"}],"version-history":[{"count":43,"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=\/wp\/v2\/pages\/851\/revisions"}],"predecessor-version":[{"id":2760,"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=\/wp\/v2\/pages\/851\/revisions\/2760"}],"up":[{"embeddable":true,"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=\/wp\/v2\/pages\/785"}],"wp:attachment":[{"href":"https:\/\/wise.ajou.ac.kr\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=851"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}