• 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) »óÈ£Á¤º¸·®°ú ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ© ±â¹ÝÀÇ ¼³¸í°¡´ÉÇÑ ¸µÅ© ¿¹Ãø
¿µ¹®Á¦¸ñ(English Title) Explainable Link Prediction based on Mutual Information and Graph Neural Networks
ÀúÀÚ(Author) Àü¼³Èñ   À̱¤Èñ   ±è¸íÈ£   Seolhee Jeon   Kwang Hee Lee   Myoung Ho Kim  
¿ø¹®¼ö·Ïó(Citation) VOL 27 NO. 09 PP. 0407 ~ 0412 (2021. 09)
Çѱ۳»¿ë
(Korean Abstract)
ÃÖ±Ù¿¡ ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©´Â ³ëµå ºÐ·ù, ±×·¡ÇÁ ºÐ·ù, ¸µÅ© ¿¹Ãø µîÀÇ ±×·¡ÇÁ ¸¶ÀÌ´×/¸Ó½Å·¯´×¿¡¼­ ÈǸ¢ÇÑ ¼º´ÉÀ» º¸¿´´Ù. ¸¹Àº ±×·¡ÇÁ ÀÓº£µù ±â¹ýµéÀÌ ±×·¡ÇÁ¿¡¼­ÀÇ ³ëµå °£ÀÇ À¯»çµµ¸¦ º¤ÅÍ °ø°£¿¡¼­ÀÇ °Å¸®·Î Ç¥ÇöÇÏ°íÀÚ ÇÑ´Ù. ½ÇÁ¦ ±×·¡ÇÁ ±¸Á¶ÀÇ µ¥ÀÌÅ͸¦ ÀÓº£µùÇÏ¿© ½ÉÃþÇнÀÀ» Çϸé ÀÌ¿ô ³ëµåµé·ÎºÎÅÍ Àü´ÞµÇ´Â Á¤º¸ÀÇ º¹ÀâÇÑ »óÈ£ÀÛ¿ë°ú ¸¹Àº ÀáÀçÀû ¿ä¼Òµé¿¡ ÀÇÇØ Á¤º¸°¡ ¾ôÈ÷±â ¶§¹®¿¡ ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©ÀÇ °á°ú¸¦ Çؼ®ÇÒ ¼ö°¡ ¾ø´Ù. ÀÌ ¹®Á¦¸¦ °³¼±Çϱâ À§ÇØ º» ³í¹®Àº ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©ÀÇ ÁÁÀº ¸µÅ© ¿¹Ãø ¼º´É°ú ¸µÅ©¿¡ ´ëÇÑ ¼³¸íÀ» Á¦½ÃÇÏ´Â »óÈ£Á¤º¸·® ±â¹ÝÀÇ ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. »óÈ£Á¤º¸·®À» È°¿ëÇÏ¿© µÎ ³ëµå °£ÀÇ À¯»çµµ¸¦ ÃÖ´ëÈ­ÇÏ´Â ·ÎÄà ¼­ºê±×·¡ÇÁ¸¦ ã°í ³ëµå °£ÀÇ ¿¬°á¿¡ ´ëÇØ Åë°èÀûÀ¸·Î ¼³¸íÇÑ´Ù. Á¦¾ÈÇÑ ¹æ¹ý¿¡¼­´Â ±âÁ¸ ±×·¡ÇÁ¿¡¼­ ¸µÅ© ¿¹Ãø¿¡ Áß¿äÇÑ ³ëµå¿Í °£¼±À» ÃßÃâÇÑ ¼­ºê±×·¡ÇÁ¸¦ ´Ù½Ã ÇнÀÇÏ¿© ¸µÅ© ¿¹ÃøÀÇ ÁÁÀº ¼º´ÉÀ» º¸ÀδÙ. µ¿½Ã¿¡ »óÈ£Á¤º¸·®¿¡ ±â¹ÝÇÏ¿© ¸µÅ© ¿¹Ãø¿¡ ´ëÇÑ ¼³¸íÀ» Á¦½ÃÇÑ´Ù.
¿µ¹®³»¿ë
(English Abstract)
Recently, graph neural networks(GNNs) have shown good performance in many graph mining/machine learning tasks such as node classification, graph classification, and link prediction. Many graph embedding techniques attempt to represent the similarity of nodes in local subgraphs as the distance in vector space. Deep learning for real-world graph-structured data poses great challenges for the interpretation of the result, as the entangled information is induced by complex interactions of neighborhood nodes together with many latent factors. To overcome this difficulty, we propose a mutual information-based method to have a good performance on link prediction as well as to produce an interpretable subgraph for the link. Mutual information is used to find a subgraph that plays an important role in maximizing the similarity of two nodes and to give an explanation statistically. Our method effectively predicts the link by utilizing subgraphs based on the importance of the nodes and the edges for link prediction. At the same time, it gives a compact explanation for link prediction by mutual information.
Å°¿öµå(Keyword) ±×·¡ÇÁ ´º·² ³×Æ®¿öÅ©   ±×·¡ÇÁ ÀÓº£µù   ¼³¸í°¡´É ÀΰøÁö´É   ¸µÅ© ¿¹Ãø   »óÈ£Á¤º¸·®   Graph Neural Networks(GNNs)   graph embedding   explainable artificial intelligence(XAI)   link prediction   mutual information  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå