• 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¸°úÇÐȸ ³í¹®Áö > Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)

Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) ÄÁº¼·ç¼Ç ´º·² ³×Æ®¿öÅ©¸¦ À§ÇÑ ¾ËÆÄ-ÀÎÅ×±×·¹ÀÌ¼Ç Ç®¸µ
¿µ¹®Á¦¸ñ(English Title) Alpha-Integration Pooling for Convolutional Neural Networks
ÀúÀÚ(Author) ¾öÇÏ¿µ   ÃÖÈñ¿­   Hayoung Eom   Heeyoul Choi  
¿ø¹®¼ö·Ïó(Citation) VOL 48 NO. 07 PP. 0774 ~ 0780 (2021. 07)
Çѱ۳»¿ë
(Korean Abstract)
ÄÁº¼·ç¼Ç ´º·² ³×Æ®¿öÅ©(CNN)´Â À̹ÌÁö ÀνÄÀ» ºñ·ÔÇÑ ¸¹Àº ¾ÖÇø®ÄÉÀ̼ǿ¡¼­ ³î¶ó¿î ¼º´ÉÀ» º¸¿©ÁÖ°í ÀÖ´Ù. CNNÀÇ ÁÖ¿ä ¿ä¼Ò Áß ÇϳªÀÎ ¼­ºê »ùÇøµÀº È¿À²ÀûÀÎ ÇнÀ°ú ºÒº¯¼º¿¡ Áß¿äÇÑ ¿ªÇÒÀ» Çϸç, ÀϹÝÀûÀ¸·Î ÃÖ´ëÇ®¸µ°ú Æò±ÕÇ®¸µÀÌ ¸¹ÀÌ »ç¿ëµÈ´Ù. µÎ ¹æ¹ý ¿Ü¿¡µµ ±âÇÏ Æò±Õ, Á¶È­ Æò±Õ µî°ú °°Àº ´Ù¸¥ Ç®¸µ À¯ÇüµéÀÌ Á¸ÀçÇÒ ¼ö ÀÖ´Ù. ¿©·¯ Ç®¸µ À¯Çüµé Áß¿¡ ÃÖÀûÀÇ À¯ÇüÀ» ÀÚµ¿À¸·Î ã±â°¡ ¾î·Æ±â ¶§¹®¿¡ ƯÁ¤ À¯ÇüÀÌ »çÀü¿¡ ¼±ÅÃµÇ¾î »ç¿ëµÇ°í ÀÌ´Â ÁÖ¾îÁø ¹®Á¦¿¡¼­ ÃÖÀûÀÇ À¯ÇüÀÌ ¾Æ´Ò ¼ö ÀÖ´Ù. ÇÏÁö¸¸, µö·¯´×ÀÇ ´Ù¸¥ º¯¼öµé°ú ¸¶Âù°¡Áö·Î, ÁÖ¾îÁø ¹®Á¦¿¡¼­ µ¥ÀÌÅͷκÎÅÍ Ç®¸µ À¯ÇüÀ» ÇнÀÇÒ ¼ö ÀÖ´Ù. º» ³í¹®Àº ÇнÀ °¡´ÉÇÑ ÆĶó¹ÌÅÍ ¥á¸¦ ÅëÇØ Ç®¸µ À¯ÇüÀ» ã¾Æ³»´Â ¾ËÆÄ-ÀÎÅ×±×·¹ÀÌ¼Ç Ç®¸µ(¥áI-pooling)À» Á¦¾ÈÇÑ´Ù. ¥áI-poolingÀº ÆĶó¹ÌÅÍ ¥á¿¡ µû¶ó ÃÖ´ëÇ®¸µ°ú Æò±ÕÇ®¸µ µîÀ» Ư¼ö ÄÉÀ̽º·Î Æ÷ÇÔÇÏ´Â ÀϹÝÈ­µÈ Ç®¸µ ¹æ¹ýÀÌ´Ù. ½ÇÇèÀ» ÅëÇØ À̹ÌÁö ÀÎ½Ä ¹®Á¦¿¡¼­ ¥áI-poolingÀÇ ¼º´ÉÀÌ ´Ù¸¥ Ç®¸µ À¯ÇüµéÀ» ´É°¡ÇÔÀ» º¸¿´´Ù. ¶ÇÇÑ, °¢°¢ ·¹À̾ ´Ù¸¥ ÃÖÀûÀÇ Ç®¸µ À¯ÇüÀ» °¡Áö°í ÀÖÀ½À» È®ÀÎÇß´Ù.
¿µ¹®³»¿ë
(English Abstract)
Convolutional neural networks (CNNs) have achieved remarkable performance in many applications, especially in image recognition tasks. As a crucial component of CNNs, sub-sampling plays an important role for efficient training or invariance property, and max-pooling and arithmetic average-pooling are commonly used sub-sampling methods. In addition to the two pooling methods, however, there are many other pooling types, such as geometric average, harmonic average, among others. Since it is not easy for algorithms to find the best pooling method, usually the pooling types are predefined, which might not be optimal for different tasks. As other parameters in deep learning, however, the type of pooling can be driven by data for a given task. In this paper, we propose ¥á-integration pooling (¥áI-pooling), which has a trainable parameter  to find the type of pooling. ¥áI-pooling is a general pooling method including max-pooling and arithmetic average-pooling as a special case, depending on the parameter ¥á. Experiments show that ¥áI-pooling outperforms other pooling methods, in image recognition tasks. Also, it turns out that each layer has a different optimal pooling type.
Å°¿öµå(Keyword) ¥áI-Pooling   ÇнÀ°¡´ÉÇÑ Ç®¸µ   ¥á-Integration   ÄÁº¼·ç¼Ç ´º·² ³×Æ®¿öÅ©   ¼­ºê »ùÇøµ   ¥áI-pooling   trainable pooling   ¥á-integration   convolutional neural networks   sub-sampling  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå