Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
·£´ý ¿öÅ© Á¡¼ö¸¦ »ç¿ëÇÑ ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ© ±â¹Ý ³ëµå ºÐ·ù |
¿µ¹®Á¦¸ñ(English Title) |
Node Classification based on Graph Neural Networks using Random Walk with Restart |
ÀúÀÚ(Author) |
¾È¼ºÁø
±è¸íÈ£
Seongjin Ahn
Myoung Ho Kim
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 26 NO. 09 PP. 0419 ~ 0423 (2020. 09) |
Çѱ۳»¿ë (Korean Abstract) |
±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©´Â ±×·¡ÇÁ¿¡¼ ³ëµå¸¦ º¤ÅÍÀÇ ÇüÅ·ΠÀÓº£µùÇÒ ¶§, ÇØ´ç ³ëµå¿Í ÀÌ¿ô °ü°è¿¡ ÀÖ´Â ³ëµåµéÀÇ Á¤º¸±îÁö ´ã¾Æ¼ ÇнÀÇÏ´Â ¹æ¹ýÀÌ´Ù. ÀϹÝÀûÀ¸·Î ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©´Â ¸ðµç ¿¬°áµÈ ³ëµåµéÀ» µ¿µîÇÑ °¡ÁßÄ¡·Î »ý°¢ÇÏ¿© Æò±ÕÀ» ³»´Â ¹æ½ÄÀ» »ç¿ëÇÔÀ¸·Î ±×·¡ÇÁ¸¦ ºÐ¼®ÇÑ´Ù. Ç㳪 ÀÌ´Â ±×·¡ÇÁÀÇ ±¸Á¶Àû ¿¬°á¼ºÀ» ¹Ý¿µÇÏÁö ¸øÇÏ¿´±â ¶§¹®¿¡ Á¤È®¼º°ú ½Å·Úµµ¸¦ ¶³¾î¶ß¸°´Ù. µû¶ó¼ º» ³í¹®Àº ·£´ý ¿öÅ©ÀÇ Á¡¼ö¸¦ ÅëÇØ ±¸Á¶Àû ¿¬°á¼º°ú À¯»ç¼ºÀ» ¾ò¾î³»°í À̸¦ ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©¿¡¼ ³ëµåÀÇ °¡ÁßÄ¡·Î ºÎ¿©ÇÔÀ¸·Î½á, ³ëµå ºÐ·ùÀÇ Á¤È®¼ºÀ» ³ôÀÌ´Â ¹æ¹ýÀ» Á¦½ÃÇÑ´Ù. ¶ÇÇÑ À̸¦ ³ëµå ºÐ·ù¿¡¼ ½ÇÇèÇÏ¿© ±âÁ¸ÀÇ ¹æ¹ýµé°ú ºñ±³ÇÑ °á°ú·Î ¿ì¸® ¹æ¹ýÀÇ ¿ì¼öÇÔÀ» º¸ÀδÙ
|
¿µ¹®³»¿ë (English Abstract) |
Graph neural networks (GNNs) are deep learning-based embedding techniques that capture the local structures and features of graphs. Traditional GNNs assume that every neighbor node has the same influence as the target node. However, each neighbor has a different influence based on its connectivity in a graph. In light of this limitation, we propose a method to increase the accuracy of GNNs by obtaining the connectivities and similarities between nodes through a random walk with restart. We also show that our method provides better accuracy in node classification tasks than existing methods
|
Å°¿öµå(Keyword) |
±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©
·£´ý ¿öÅ©
±×·¡ÇÁ ÀÓº£µù
³ëµå ºÐ·ù
graph neural network
random walk
graph embedding
node classification
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|