• 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¸Åë½ÅÇÐȸ ³í¹®Áö (Journal of the Korea Institute of Information and Communication Engineering)

Çѱ¹Á¤º¸Åë½ÅÇÐȸ ³í¹®Áö (Journal of the Korea Institute of Information and Communication Engineering)

Current Result Document : 10 / 17 ÀÌÀü°Ç ÀÌÀü°Ç   ´ÙÀ½°Ç ´ÙÀ½°Ç

ÇѱÛÁ¦¸ñ(Korean Title) °¡¼Ó ȸ·Î¿¡ ÀûÇÕÇÑ CNNÀÇ Conv-XP °¡ÁöÄ¡±â
¿µ¹®Á¦¸ñ(English Title) Conv-XP Pruning of CNN Suitable for Accelerator
ÀúÀÚ(Author) ¿ì¿ë±Ù   °­ÇüÁÖ   Yonggeun Woo   Hyeong-Ju Kang  
¿ø¹®¼ö·Ïó(Citation) VOL 23 NO. 01 PP. 0055 ~ 0062 (2019. 01)
Çѱ۳»¿ë
(Korean Abstract)
CNNÀº ÄÄÇ»ÅÍ ¿µ»ó ÀÎ½Ä ºÎºÐ¿¡¼­ ³ôÀº ¼º´ÉÀ» º¸¿©ÁÖ°í ÀÖÀ¸³ª ¸¹Àº ¿¬»ê¾çÀ» ¿ä±¸ÇÏ´Â ´ÜÁ¡À¸·Î ÀÎÇØ Àü·ÂÀ̳ª ¿¬»ê ´É·Â¿¡ Á¦ÇÑÀÌ ÀÖ´Â ÀÓº£µðµå ȯ°æ¿¡¼­´Â »ç¿ëÇϱ⠾î·Æ´Ù. ÀÌ·¯ÇÑ ´ÜÁ¡À» ±Øº¹Çϱâ À§ÇØ CNNÀ» À§ÇÑ °¡¼Óȸ·Î³ª °¡ÁöÄ¡±â ±â¹ý¿¡ ´ëÇÑ ¿¬±¸°¡ ¸¹ÀÌ ÀÌ·ç¾îÁö°í ÀÖ´Ù. ±âÁ¸ÀÇ °¡ÁöÄ¡±â ±â¹ýÀº °¡¼Ó ȸ·ÎÀÇ ±¸Á¶¸¦ °í·ÁÇÏÁö ¾Ê¾Æ¼­, °¡ÁöÄ¡±âµÈ CNNÀ» À§ÇÑ °¡¼Ó ȸ·Î´Â ºñÈ¿À²ÀûÀÎ ±¸Á¶¸¦ °¡Áö°Ô µÈ´Ù. ÀÌ ³í¹®¿¡¼­´Â °¡¼Ó ȸ·ÎÀÇ ±¸Á¶¸¦ °í·ÁÇÑ »õ·Î¿î °¡ÁöÄ¡±â ±â¹ýÀÎ Conv-XP °¡ÁöÄ¡±â¸¦ Á¦¾ÈÇÑ´Ù. Conv-XP °¡ÁöÄ¡±â¿¡¼­´Â ¡®X¡¯¿Í ¡®+¡¯ ¸ð¾çÀÇ µÎ °¡Áö ÆÐÅÏÀ¸·Î¸¸ °¡ÁöÄ¡±âÇÔÀ¸·Î½á, ÀÌ ±â¹ýÀ¸·Î °¡ÁöÄ¡±âµÈ CNNÀ» À§ÇÑ °¡¼Ó ȸ·ÎÀÇ ±¸Á¶¸¦ ´Ü¼øÇÏ°Ô ¼³°èÇÒ ¼ö ÀÖµµ·Ï ÇÏ¿´´Ù. ½ÇÇè °á°ú¿¡ µû¸£¸é, Conv-XP¿Í °°ÀÌ °¡ÁöÄ¡±â ÆÐÅÏÀ» Á¦ÇÑÇÏ¿©µµ CNNÀÇ ¼º´ÉÀÌ ¾ÇÈ­µÇÁö ¾ÊÀ¸¸ç, °¡¼Ó ȸ·ÎÀÇ ¸éÀûÀº 12.8%À» °¨¼Ò½Ãų ¼ö ÀÖ´Ù.
¿µ¹®³»¿ë
(English Abstract)
Convolutional neural networks (CNNs) show high performance in the computer vision, but they require an enormous amount of operations, making them unsuitable for some resource- or energy-starving environments like the embedded environments. To overcome this problem, there have been much research on accelerators or pruning of CNNs. The previous pruning schemes have not considered the architecture of CNN accelerators, so the accelerators for the pruned CNNs have some inefficiency. This paper proposes a new pruning scheme, Conv-XP, which considers the architecture of CNN accelerators. In Conv-XP, the pruning is performed following the ¡®X¡¯ or ¡® ¡¯ shape. The Conv-XP scheme induces a simple architecture of the CNN accelerators. The experimental results show that the Conv-XP scheme does not degrade the accuracy of CNNs, and that the accelerator area can be reduced by 12.8%.
Å°¿öµå(Keyword) ½Å°æ¸Á   CNN   °¡¼Ó ȸ·Î   °¡ÁöÄ¡±â   Neural networks   Convolutional neural networks   Accelerator   Pruning  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå