Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)
ÇѱÛÁ¦¸ñ(Korean Title) |
°°ÇÇÑ ½ºÄÉÁÙ¸µÀ» À§ÇÑ ¸¶ÄÚÇÁ ÀÇ»ç°áÁ¤ ÇÁ·Î¼¼½º Ãß·Ð ¹× ¿ª°È ÇнÀ ±â¹Ý ÀÏ»ó Çൿ ÇнÀ |
¿µ¹®Á¦¸ñ(English Title) |
Robust Scheduling based on Daily Activity Learning by using Markov Decision Process and Inverse Reinforcement Learning |
ÀúÀÚ(Author) |
ÀÌ»ó¿ì
°ûµ¿Çö
¿Â°æ¿î
ÇãÀ¯Á¤
°¿ì¿µ
ÀçÀÌ´Ù
À庴Ź
Sang-Woo Lee
Dong-Hyun Kwak
Kyoung-Woon On
Yujung Heo
Wooyoung Kang
Ceyda Cinarel
Byoung-Tak Zhang
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 23 NO. 10 PP. 0599 ~ 0604 (2017. 10) |
Çѱ۳»¿ë (Korean Abstract) |
À¯ÀúÀÇ ÀÏ»ó ½ºÄÉÁìÀ» Á¦¾ÈÇÏ°í ¿¹ÃøÇÏ´Â ¼ºñ½º´Â ½º¸¶Æ® ºñ¼ÀÇ Èï¹Ì·Î¿î ÀÀ¿ëÀÌ´Ù. ÀüÅëÀûÀÎ ¹æ¹ý¿¡¼´Â À¯ÀúÀÇ ÇൿÀ» ¿¹ÃøÇϱâ À§ÇÏ¿©, À¯Àú°¡ Á÷Á¢ ÀÚ½ÅÀÇ ÇൿÀ» ±â·ÏÇϰųª, e-mail ȤÀº SNS µî¿¡¼ ¸í½ÃÀûÀÎ ÀÏÁ¤ Á¤º¸¸¦ ÃßÃâÇÏ¿© »ç¿ëÇØ¿Ô´Ù. ÇÏÁö¸¸, À¯Àú°¡ ¸ðµç Á¤º¸¸¦ ±â·ÏÇÒ ¼ö ¾ø±â¿¡, ½º¸¶Æ® ºñ¼°¡ ¾òÀ» ¼ö ÀÖ´Â Á¤º¸´Â Á¦ÇÑÀûÀ̸ç, À¯Àú´Â À¯ÀúÀÇ ÀÏ»óÀÇ routineÇÑ Á¤º¸¸¦ ±â·ÏÇÏÁö ¾Ê´Â °æÇâÀÌ ÀÖ´Ù. º» ³í¹®¿¡¼´Â ½ºÄÉÁÙ·¯¿¡ ÀûÈ÷´Â Á¤ÇüÈµÈ ÀÏÁ¤ÀÎ ½ºÄÉÁÙ°ú ºñÁ¤ÇüÈµÈ ÀÏÁ¤À» ¸¸µå´Â ÀÏ»ó Çൿ ÆÐÅϵéÀ» µ¿½Ã¿¡ °í·ÁÇÏ´Â Á¢±Ù ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. À̸¦ À§ÇÏ¿© ¸¶ÄÚÇÁ ÀÇ»ç °áÁ¤ ÇÁ·Î¼¼½º (MDP)¸¦ ±â¹ÝÀ¸·Î ÇÏ´Â Ãß·Ð ¹æ¹ý°ú ¿ª°È ÇнÀ (IRL)À» ÅëÇÑ º¸»ó ÇÔ¼ö ÇнÀ ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. ½ÇÇè °á°ú´Â ¿ì¸®°¡ 6ÁÖ°£ ¸ðÀº ½ÇÁ¦ »ýÈ°À» ±â·ÏÇÑ µ¥ÀÌÅÍ ¼Â¿¡¼ ¿ì¸®ÀÇ ¹æ¹ýÀÌ ±âÁ¸ ¹æ¹ýµéº¸´Ù ¿ì¼öÇÑ ¼º´ÉÀ» º¸ÀÓÀ» ³íÁõÇÑ´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
A useful application of smart assistants is to predict and suggest users¡¯ daily behaviors the way real assistants do. Conventional methods to predict behavior have mainly used explicit schedule information logged by a user or extracted from e-mail or SNS data. However, gathering explicit information for smart assistants has limitations, and much of a user¡¯s routine behavior is not logged in the first place. In this paper, we suggest a novel approach that combines explicit schedule information with patterns of routine behavior. We propose using inference based on a Markov decision process and learning with a reward function based on inverse reinforcement learning. The results of our experiment shows that the proposed method outperforms comparable models on a life-log dataset collected over six weeks.
|
Å°¿öµå(Keyword) |
ÀÏÁ¤ °ü¸®
¸¶ÄÚÇÁ ÀÇ»ç °áÁ¤ ÇÁ·Î¼¼½º
¿ª°È ÇнÀ
¿þ¾î·¯ºí µð¹ÙÀ̽º
schedule planning
markov decision process
inverse reinforcement learning
wearable devices
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|