BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035311Z
LOCATION:Room 323\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221208T124500
DTEND;TZID=Asia/Seoul:20221208T130000
UID:siggraphasia_SIGGRAPH Asia 2022_sess258_gp_134@linklings.com
SUMMARY:A Visual Analytics System for Improving Attention-based Traffic Fo
 recasting Models
DESCRIPTION:Talks\n\nA Visual Analytics System for Improving Attention-bas
 ed Traffic Forecasting Models\n\nLee,\n\nWith deep learning (DL) outperfor
 ming conventional methods for different tasks, much effort has been devote
 d to utilizing DL in various domains. Researchers and developers in the tr
 affic domain have also designed and improved DL models for forecasting tas
 ks such as estimation of traffic speed and time of arrival. However, there
  exist many challenges in analyzing DL models due to the black-box propert
 y of DL models and complexity of traffic data (i.e., spatio-temporal depen
 dencies). Collaborating with domain experts, we design a visual analytics 
 system, AttnAnalyzer, that enables users to explore how DL models make pre
 dictions by allowing effective spatio-temporal dependency analysis. The sy
 stem incorporates dynamic time warping (DTW) and Granger causality tests f
 or computational spatio-temporal dependency analysis while providing map, 
 table, line chart, and pixel views to assist user to perform dependency an
 d model behavior analysis. For the evaluation, we present three case studi
 es showing how AttnAnalyzer can effectively explore model behaviors and im
 prove model performance in two different road networks. We also provide do
 main expert feedback.\n\nAuthors: Seungmin Jin, Hyunwook Lee, Cheonbok Par
 k, Hyeshin Chu, Yunwon Tae, Jaegul Choo, Sungahn Ko\n\nRegistration Catego
 ry: FULL ACCESS, EXPERIENCE PLUS ACCESS, TRADE EXHIBITOR\n\nLanguage: ENGL
 ISH\n\nFormat: IN-PERSON
URL:https://sa2022.siggraph.org/en/full-program/?id=gp_134&sess=sess258
END:VEVENT
END:VCALENDAR