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

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ³í¹®Áö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ³í¹®Áö > Çѱ¹Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö > Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö ¼ÒÇÁÆ®¿þ¾î ¹× µ¥ÀÌÅÍ °øÇÐ

Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö ¼ÒÇÁÆ®¿þ¾î ¹× µ¥ÀÌÅÍ °øÇÐ

Current Result Document : 468 / 469

ÇѱÛÁ¦¸ñ(Korean Title) ½ºÅµ¿¬°áÀÌ Àû¿ëµÈ ¿ÀÅäÀÎÄÚ´õ ¸ðµ¨ÀÇ Å¬·¯½ºÅ͸µ ¼º´É ºÐ¼®
¿µ¹®Á¦¸ñ(English Title) Clustering Performance Analysis of Autoencoder with Skip Connection
ÀúÀÚ(Author) Á¶Àμö   °­À±Èñ   ÃÖµ¿ºó   ¹Ú¿ë¹ü   In-su Jo   Yunhee Kang   Dong-bin Choi   Young B. Park  
¿ø¹®¼ö·Ïó(Citation) VOL 09 NO. 12 PP. 0403 ~ 0410 (2020. 12)
Çѱ۳»¿ë
(Korean Abstract)
¿ÀÅäÀÎÄÚ´õÀÇ µ¥ÀÌÅÍ º¹¿ø(Output result) ±â´ÉÀ» ÀÌ¿ëÇÑ ³ëÀÌÁî Á¦°Å ¹× ÃÊÇØ»óµµ¿Í °°Àº ¿¬±¸°¡ ÁøÇàµÇ´Â °¡¿îµ¥ ¿ÀÅäÀÎÄÚ´õÀÇ Â÷¿ø Ãà¼Ò ±â´ÉÀ» ÀÌ¿ëÇÑ Å¬·¯½ºÅ͸µÀÇ ¼º´É Çâ»ó¿¡ ´ëÇÑ ¿¬±¸µµ È°¹ßÈ÷ ÁøÇàµÇ°í ÀÖ´Ù. ¿ÀÅäÀÎÄÚ´õ¸¦ ÀÌ¿ëÇÑ Å¬·¯½ºÅ͸µ ±â´É°ú µ¥ÀÌÅÍ º¹¿ø ±â´ÉÀº ¸ðµÎ µ¿ÀÏÇÑ ÇнÀÀ» ÅëÇØ ¼º´ÉÀ» Çâ»ó½ÃŲ´Ù´Â °øÅëÁ¡ÀÌ ÀÖ´Ù. º» ³í¹®Àº ÀÌ·± Ư¡À» Åä´ë·Î, µ¥ÀÌÅÍ º¹¿ø ¼º´ÉÀÌ ¶Ù¾î³ªµµ·Ï ¼³°èµÈ ¿ÀÅäÀÎÄÚ´õ ¸ðµ¨ÀÌ Å¬·¯½ºÅ͸µ ¼º´É ¶ÇÇÑ ¶Ù¾î³­Áö ¾Ë¾Æº¸±â À§ÇÑ ½ÇÇèÀ» ÁøÇàÇß´Ù. µ¥ÀÌÅÍ º¹¿ø ¼º´ÉÀÌ ¶Ù¾î³­ ¿ÀÅäÀÎÄÚ´õ¸¦ ¼³°èÇϱâ À§Çؼ­ ½ºÅµ¿¬°á(Skip connection) ±â¹ýÀ» »ç¿ëÇß´Ù. ½ºÅµ¿¬°á ±â¹ýÀº ±â¿ï±â ¼Ò½Ç(Vanishing gradient)Çö»óÀ» ÇؼÒÇØÁÖ°í ¸ðµ¨ÀÇ ÇнÀ È¿À²À» ³ôÀδٴ ÀåÁ¡À» °¡Áö°í ÀÖÀ» »Ó¸¸ ¾Æ´Ï¶ó, µ¥ÀÌÅÍ º¹¿ø ½Ã ¼Õ½ÇµÈ Á¤º¸¸¦ º¸¿ÏÇØ ÁÜÀ¸·Î½á µ¥ÀÌÅÍ º¹¿ø ¼º´ÉÀ» ³ôÀÌ´Â È¿°úµµ °¡Áö°í ÀÖ´Ù. ½ºÅµ¿¬°áÀÌ Àû¿ëµÈ ¿ÀÅäÀÎÄÚ´õ ¸ðµ¨°ú Àû¿ëµÇÁö ¾ÊÀº ¸ðµ¨ÀÇ µ¥ÀÌÅÍ º¹¿ø ¼º´É°ú Ŭ·¯½ºÅ͸µ ¼º´ÉÀ» ±×·¡ÇÁ¿Í ½Ã°¢Àû ÃßÃâ¹°À» ÅëÇØ °á°ú¸¦ ºñ±³ÇØ º¸´Ï, µ¥ÀÌÅÍ º¹¿ø ¼º´ÉÀº ¿Ã¶úÁö¸¸ Ŭ·¯½ºÅ͸µ ¼º´ÉÀº ¶³¾îÁö´Â °á°ú¸¦ È®ÀÎÇß´Ù. ÀÌ °á°ú´Â ¿ÀÅäÀÎÄÚ´õ¿Í °°Àº ½Å°æ¸Á ¸ðµ¨ÀÌ Ãâ·ÂµÈ °á°ú ¼º´ÉÀÌ ÁÁ´Ù°í Çؼ­ °¢ ·¹À̾îµéÀÌ µ¥ÀÌÅÍÀÇ Æ¯Â¡À» ¸ðµÎ Àß ÇнÀÇß´Ù°í È®½ÅÇÒ ¼ö ¾øÀ½À» ¾Ë·ÁÁØ´Ù. ¸¶Áö¸·À¸·Î Ŭ·¯½ºÅ͸µÀÇ ¼º´ÉÀ» Á¿ìÇÏ´Â ÀáÀ纯¼ö(latent code)¿Í ½ºÅµ¿¬°áÀÇ °ü°è¸¦ ºÐ¼®ÇÏ¿© ½ÇÇè °á°úÀÇ ¿øÀο¡ ´ëÇØ ÆľÇÇÏ¿´°í, ÆľÇÇÑ °á°ú¸¦ ÅëÇØ ÀáÀ纯¼ö¿Í ½ºÅµ¿¬°áÀÇ Æ¯Â¡Á¤º¸¸¦ ÀÌ¿ëÇØ Å¬·¯½ºÅ͸µÀÇ ¼º´ÉÀúÇÏ Çö»óÀ» º¸¿ÏÇÒ ¼ö ÀÖ´Ù´Â »ç½ÇÀ» º¸¿´´Ù. ÀÌ ¿¬±¸´Â ÇÑÀÚ À¯´ÏÄÚµå ¹®Á¦¸¦ Ŭ·¯½ºÅ͸µ ±â¹ýÀ» ÀÌ¿ëÇØ ÇØ°áÇÏ°íÀÚ Å¬·¯½ºÅ͸µ ¼º´É Çâ»óÀ» À§ÇÑ ¼±Ç࿬±¸ÀÌ´Ù.
¿µ¹®³»¿ë
(English Abstract)
In addition to the research on noise removal and super-resolution using the data restoration (Output result) function of Autoencoder, research on the performance improvement of clustering using the dimension reduction function of autoencoder are actively being conducted. The clustering function and data restoration function using Autoencoder have common points that both improve performance through the same learning. Based on these characteristics, this study conducted an experiment to see if the autoencoder model designed to have excellent data recovery performance is superior in clustering performance. Skip connection technique was used to design autoencoder with excellent data recovery performance. The output result performance and clustering performance of both autoencoder model with Skip connection and model without Skip connection were shown as graph and visual extract. The output result performance was increased, but the clustering performance was decreased. This result indicates that the neural network models such as autoencoders are not sure that each layer has learned the characteristics of the data well if the output result is good. Lastly, the performance degradation of clustering was compensated by using both latent code and skip connection. This study is a prior study to solve the Hanja Unicode problem by clustering
Å°¿öµå(Keyword) ½ºÅµ¿¬°á   ¿ÀÅäÀÎÄÚ´õ   Ŭ·¯½ºÅ͸µ   ÃÊÇػ󵵠  Skip Connection   Autoencoder   Clustering   Superresolution  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå