• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ÇÐȸÁö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ÇÐȸÁö > µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) ±â¾÷Çü ¿¬ÇÕÇнÀ ÇÁ·¹ÀÓ¿öÅ© ºÐ¼®
¿µ¹®Á¦¸ñ(English Title) An Analysis on Federated Learning Frameworks for Enterprises
ÀúÀÚ(Author) ±èä¹Ì   Chaemee Kim   ³ë¿õ±â   Woong-Kee Loh  
¿ø¹®¼ö·Ïó(Citation) VOL 36 NO. 02 PP. 0028 ~ 0044 (2020. 08)
Çѱ۳»¿ë
(Korean Abstract)
µ¥ÀÌÅ͸¦ ÇÑ°÷¿¡ ¸ð¾Æ¼­ ÇнÀÇÏ´Â ±âÁ¸ÀÇ ¹æ½ÄÀº µ¥ÀÌÅÍÀÇ ¼öÁý ´Ü°èºÎÅÍ ¾î·Á¿ò¿¡ Á÷¸éÇϱ⠽±´Ù. ¹Î°¨ÇÑ °³ÀÎ Á¤º¸¿Í ¼­ºñ½º µ¥ÀÌÅÍÀÇ À¯ÃâÀ̳ª À̵¿¿¡ ´ëÇؼ­ Á¦¾àÀÌ ¸¹°í, ¾Ïȣȭ¿Í ºñ½Äº°È­¿Í °°Àº Á¤º¸º¸¾È ´ëÃ¥À» Àû¿ëÇصµ µ¥ÀÌÅÍ°¡ ÁýÁßÀûÀ¸·Î ¸ð¿©ÀÖ´Â °÷Àº º¸¾È¿¡ Ãë¾àÇÑ È¯°æÀÏ ¼ö¹Û¿¡ ¾ø´Ù. ¿¬ÇÕÇнÀ(federated learning)Àº µ¥ÀÌÅÍ°¡ À̵¿ÇÏ´Â ´ë½Å µ¥ÀÌÅÍ°¡ ÀÖ´Â ´Ù¾çÇÑ °÷¿¡¼­ ÇнÀÀÌ ÀÌ·ç¾îÁö°í, °¢°¢ÀÇ ÇнÀ°á°ú »ý¼ºµÈ °³º° ·ÎÄà ÇнÀ¸ðµ¨(local learning model)µéÀ» ÅëÇÕÇÑ ±Û·Î¹ú ÇнÀ¸ðµ¨(global learning model)À» ´Ù½Ã µ¥ÀÌÅÍ°¡ ÀÖ´Â °÷¿¡ ¹èÄ¡ÇÏ´Â ¹æ½ÄÀÌ´Ù. ¿¬ÇÕÇнÀÀº ¼ö¸¹Àº ´Ü¸»±â ¶Ç´Â ºÐ»ê ¸Ó½Å¿¡¼­ »ý¼ºµÈ µ¥ÀÌÅÍ¿Í °¡±î¿î °÷¿¡¼­ ó¸®ÇÏ°í È°¿ëÇÏ´Â ¿¡Áö ÄÄÇ»ÆÃ(edge computing), ±â¾÷ ³»ºÎ¶óµµ µ¥ÀÌÅÍ À̵¿ÀÇ Á¦¾àÀÌ ÀÖ´Â ¼­ºñ½º °£ÀÇ µ¥ÀÌÅÍ È°¿ë, µ¿ÀÏÇÑ »ê¾÷±º ³»¿¡¼­ µ¥ÀÌÅÍ °øÀ¯´Â ¾î·ÆÁö¸¸ À¶ÇÕ ¼­ºñ½ºÀÇ Çʿ伺ÀÌ ÀÖ´Â ÀÀ¿ë ¿¡ Àû¿ëÇÒ ¼ö Àִ ȹ±âÀûÀÎ ¸Ó½Å·¯´× ¹æ½ÄÀÌ´Ù. º» ³í¹®¿¡¼­´Â ´Ù¾çÇÑ ±â¾÷ ÄÄÇ»Æà ȯ°æ¿¡¼­ ÀûÇÕÇÑ ¿¬ÇÕ¸ðµ¨À» È¿À²ÀûÀ¸·Î ¿î¿µÇϱâ À§ÇÑ ±âÁ¸ÀÇ ¿¬±¸µéÀ» ¼Ò°³ÇÏ°í, ±â¾÷Çü ¿¬ÇÕÇнÀ ÇÁ·¹ÀÓ¿öÅ©·Î ¹ßÀü½ÃÅ°±â À§ÇÑ ¹®Á¦Á¡°ú ÇâÈÄ ¿¬±¸ ¹æÇâÀ» Á¦½ÃÇÑ´Ù. ±â¾÷Çü ¿¬ÇÕÇнÀÀº ±â¾÷ °£(B2B), ±â¾÷°ú °³ÀÎ °£(B2C) µ¥ÀÌÅÍÀÇ À̵¿ ¾øÀÌ ´Ù¾çÇÑ ¸Ó½Å·¯´× ÀÀ¿ëÀ» °¡´ÉÇÏ°Ô ÇÑ´Ù´Â Á¡¿¡¼­ µ¥ÀÌÅÍ ÁÖµµÇü »ê¾÷ º¯È­¸¦ ¼±µµÇÒ ¼ö ÀÖ´Â ÀáÀçÀû ¼ºÀ强ÀÌ ³ôÀº ¿¬±¸°¡ µÉ °ÍÀÌ´Ù.
¿µ¹®³»¿ë
(English Abstract)
The traditional machine learning approach that performs training using massive data in a central server tends to face difficulties from the beginning of data collection. Most sensitive private data and service business data are restricted to flow outside, and even with the data protection arrangements such as encryption and de-identification, the data collection sites are major targets of security breach. In federated learning, instead of collecting data, learning algorithms are run at each of various servers and devices where data are stored, and the resulting models are sent to a central server to build a global integrated model, which is then deployed back to each data server and device. Federated learning is an innovative approach that can be adopted in various areas including (1) an edge computing system where data storage, computation, and application are carried out at close locations using mobile devices and distributed servers, (2) application of highly flow-restrictive intra-enterprise service data, and (3) integration of intra-industry services using hardly sharable data. In this paper, we introduce previous research work for efficiently adopting and managing federated learning models that are adequate for various enterprise computing environments and present the issues and future research directions for improvement of enterprise federated learning frameworks. We believe that enterprise federated learning will solve many machine learning problems in B2B and B2C environments without data movements and therefore play a crucial role of leading data-centric industrial innovations.
Å°¿öµå(Keyword) ¿¬ÇÕÇнÀ   ±â¾÷Çü ÇÁ·¹ÀÓ¿öÅ©   ¿¡Áö ÄÄÇ»Æà  ºÐ»ê ¸Ó½Å·¯´×   ¸¶À̵¥ÀÌÅÍ   federated learning   enterprise framework   edge computing   distributed machine learning   MyData  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå