• 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¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) Àΰø½Å°æ¸ÁÀÇ È¿À²Àû ³ëµå »çÈÄ °¡ÁöÄ¡±â
¿µ¹®Á¦¸ñ(English Title) Efficient Node Posterior Pruning with Artificial Neural Networks
ÀúÀÚ(Author) À±½ÂÇÑ   ÀÌÀ翵   ÀÌ¿õÈñ   ±è¿µÈÆ   Seunghan Yoon   Jaeyoung Lee   Woonghee Lee   Younghoon Kim  
¿ø¹®¼ö·Ïó(Citation) VOL 28 NO. 03 PP. 0184 ~ 0189 (2022. 03)
Çѱ۳»¿ë
(Korean Abstract)
½ÉÃþ½Å°æ¸ÁÀº À̹ÌÁö ºÐ·ù ¹®Á¦¿¡¼­ ³ôÀº Á¤È®µµ¸¦ º¸À̸ç ÁÖ¸ñÀ» ¹Þ¾Ò´Ù. ÇÏÁö¸¸ ºÐ·ùÇØ¾ß ÇÏ´Â ÇнÀµ¥ÀÌÅÍ¿Í Å¬·¡½º°¡ ¸¹¾ÆÁö°í º¹ÀâÇØÁö¸é ¸¹Àº ¸Å°³º¯¼ö¸¦ ÇÊ¿ä·Î Çϱ⠶§¹®¿¡ ³×Æ®¿öÅ©ÀÇ Å©±â°¡ Ä¿Áö°í ¿¬»ê·®µµ ¸¹¾ÆÁø´Ù. ±×·¡¼­ ºÐ·ù ¼º´ÉÀº À¯ÁöÇϸ鼭 ³×Æ®¿öÅ©ÀÇ Å©±â¸¦ ÁÙÀÌ´Â ¿¬±¸°¡ ¸¹ÀÌ ÁøÇàµÇ°í ÀÖ´Ù. º» ¿¬±¸´Â À̹ÌÁö ºÐ·ù¸¦ ¼öÇàÇÏ´Â ½ÉÃþ½Å°æ¸ÁÀÇ ¿ÏÀü¿¬°á°èÃþ¿¡¼­ °¡ÁöÄ¡±â¸¦ ¼öÇàÇÏ´Â ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. ºÐ·ù¿¡ ¸¹Àº ¿µÇâÀ» ¹ÌÄ¡´Â ³ëµåÁýÇÕÀ» °ËÃâÇÏ¿© ±× ³ëµåµé¸¸À» Ãß·Ð ÀÛ¾÷¿¡ È°¿ëÇÑ´Ù. ÃßÃâµÈ Ư¡ÀÌ ¸ðÀÌ´Â Ãâ·Â°èÃþÀÇ Á÷Àü°èÃþµé¿¡¼­ °¡ÁöÄ¡±â¸¦ ¼öÇàÇÏ°í ±âÁ¸ ¿¬±¸¿Í ºÐ·ù Á¤È®µµ¸¦ ºñ±³ÇÏ¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Deep neural networks have attracted attention with high accuracy in image classification problems. However, as the number of training data and classes to be classified increases and becomes more complex, the size of the network and the amount of computation increase because many parameters are required. Thus, many studies are being conducted to reduce the size of the network while maintaining the classification performance. This study proposes a method to perform pruning at the fully connected layer of deep neural networks that perform image classification. A set of nodes that have considerable influence on classification are detected, and only those nodes are used for inference. Pruning was performed in the layers immediately preceding the output layer where the extracted features were gathered, and the classification accuracy was compared with the previous studies.
Å°¿öµå(Keyword) ½ÉÃþ½Å°æ¸Á   ¿ÏÀü¿¬°á°èÃþ   ½Å°æ¸Á °¡ÁöÄ¡±â   À̹ÌÁö ºÐ·ù   ½ÉÃþ½Å°æ¸Á   ¿ÏÀü¿¬°á°èÃþ   ½Å°æ¸Á °¡ÁöÄ¡±â   À̹ÌÁö ºÐ·ù   LightGBM   XAI   solar power generation   time series forecasting  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå