BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180035Z
LOCATION:3001\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240625T103000
DTEND;TZID=America/Los_Angeles:20240625T104500
UID:dac_DAC 2024_sess103_RESEARCH1015@linklings.com
SUMMARY:Graph Neural Networks Automated Design and Deployment on Device-Ed
 ge Co-Inference Systems
DESCRIPTION:Research Manuscript\n\nAo Zhou, Jianlei Yang, Tong Qiao, and Y
 ingjie Qi (Beihang University); Zhi Yang (Peking University); and Weisheng
  Zhao and Chunming Hu (Beihang University)\n\nThe key to device-edge co-in
 ference paradigm is to partition models into computation-friendly and comp
 utation-intensive parts across device and edge, respectively. However, for
  Graph Neural Networks (GNNs), partitioning without architecture explorati
 on is ineffective due to various computational-communication overheads of 
 GNN operations over heterogeneous devices. We present GCoDE, the first aut
 omatic framework that co-designs the GNN architecture and operation mappin
 g. GCoDE abstracts communication process into explicit operation, fuses ar
 chitecture search and operations mapping in a joint-optimization space. Al
 so, the performance-awareness approach enables effective evaluation of arc
 hitecture efficiency. Experiments show GCoDE achieves up to 44.9x speedup 
 and 98.2% energy reduction across various systems.\n\nTopic: AI\n\nKeyword
 : AI/ML Algorithms\n\nSession Chairs: Hongyang Jia (Tsinghua University) a
 nd Grace Li Zhang (Technische Universität Darmstadt)
END:VEVENT
END:VCALENDAR