KCC 2021
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
Àü´Þ¼º Ä¿³ÎÀ» »ç¿ëÇÑ ±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ© ±â¹Ý °£¼± ¿¹Ãø |
¿µ¹®Á¦¸ñ(English Title) |
Link Prediction Based on Graph Convolutional Networks using Communicability Kernel |
ÀúÀÚ(Author) |
¾È¼ºÁø
±è¸íÈ£
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 48 NO. 01 PP. 0783 ~ 0785 (2021. 06) |
Çѱ۳»¿ë (Korean Abstract) |
ÃÖ±Ù¿¡ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©¸¦ ±×·¡ÇÁ ºÐ¾ß·Î È®ÀåÇÑ ±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©´Â ³ëµå ºÐ·ù, ±×·¡ÇÁ ºÐ·ù, ¸µÅ© ¿¹Ãø µîÀÇ ¿©·¯ ±×·¡ÇÁ ºÐ¼® ¹®Á¦¿¡¼ ÈǸ¢ÇÑ ¼º´ÉÀ» º¸¿´´Ù. ÇÏÁö¸¸ ±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©µéÀº °¢ ³ëµåÀÇ ÀÎÁ¢ÇÑ ±¸Á¶¸¸À» ÆľÇÇÏ¿© ³ëµå¸¦ ÀúÂ÷¿ø º¤ÅͷΠǥÇöÇÏ¿© ÇнÀÇϴ Ư¼ºÀÌ ÀÖ´Ù. ÀüüÀûÀÎ ±¸Á¶¸¦ ¹Ý¿µÇÏ·Á´Â °íÂ÷ ±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©µéÀÌ Á¦½ÃµÇ¾ú´Ù. ÇÏÁö¸¸ ±âÁ¸ÀÇ °íÂ÷ ±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©µéÀº ±×·¡ÇÁÀÇ °£¼±À» ¿¹ÃøÇϴµ¥¿¡´Â ÁÁÀº ¼º´ÉÀ» º¸ÀÌÁö ¸øÇÏ°í ÀÖ´Ù. º» ³í¹®¿¡¼´Â ±âÁ¸ÀÇ °íÂ÷ ±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©µéÀÌ ±æÀÌ°¡ ´Ù¸¥ °æ·Î¿¡¼ ¾ò¾îÁø Á¤º¸µéÀ» Â÷µîÀûÀ¸·Î ¹Þ¾ÆµéÀÌÁö ¸øÇÏ´Â ¹®Á¦¸¦ ÁöÀûÇÑ´Ù. ±×¸®°í ÀÌ·¯ÇÑ ÇѰ踦 º¸¿ÏÇϱâ À§Çؼ, Àü´Þ¼º Ä¿³ÎÀ» È°¿ëÇÏ¿© °£¼± ¿¹Ãø¿¡ ¾Ë¸Â°Ô Àüü ±×·¡ÇÁÀÇ ±¸Á¶¸¦ ¹Ý¿µÇÑ °íÂ÷ ±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©¸¦ Á¦½ÃÇÑ´Ù. |
¿µ¹®³»¿ë (English Abstract) |
Recently, Graph Convolutional Networks (GCNs) which expand convolutional neural networks to the graph fields have shown good performance in various graph analysis problems such as the node classification, the graph classification, and the link prediction. GCNs generate node embeddings by using features of the corresponding node and its 1-hop neighbors. Higher-order graph convolutional networks (Higher-order GCNs) have been proposed in order to utilize the information from global structure. However, existing higher-order GCNs do not predict links properly. We find out that considering different lengths of paths makes higher-order GCNs predict links better. In this paper, we propose a Communicability Graph Convolutional Networks (CGCN) that assigns different weights to the length of the path by using a communicability kernel. |
Å°¿öµå(Keyword) |
±×·¡ÇÁ ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©
±×·¡ÇÁ ÀÓº£µù
¸µÅ© ¿¹Ãø
Àü´Þ¼º Ä¿³Î
Graph Convolutional Network
Graph Embedding
Link Prediction
Communicability Kernel
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|