BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T000713Z
LOCATION:301-302-303
DTSTART;TZID=America/Denver:20231114T163000
DTEND;TZID=America/Denver:20231114T170000
UID:submissions.supercomputing.org_SC23_sess167_pap273@linklings.com
SUMMARY:DistTGL: Distributed Memory-Based Temporal Graph Neural Network Tr
 aining
DESCRIPTION:Hongkuan Zhou (University of Southern California (USC)); Da Zh
 eng, Xiang Song, and George Karypis (Amazon Web Services AI); and Viktor P
 rasanna (University of Southern California (USC))\n\nMemory-based Temporal
  Graph Neural Networks are powerful tools in dynamic graph representation 
 learning and have demonstrated superior performance in many real-world app
 lications.  However, their node memory favors smaller batch sizes to captu
 re more dependencies in graph events and needs to be maintained synchronou
 sly across all trainers.  As a result, existing frameworks suffer from acc
 uracy loss when scaling to multiple GPUs. Even worse, the tremendous overh
 ead to synchronize the node memory make it impractical to be deployed to d
 istributed GPU clusters. \n\nIn this work, we propose DistTGL --- an effic
 ient and scalable solution to train memory-based TGNNs on distributed GPU 
 clusters.\nDistTGL has three improvements over existing solutions: an enha
 nced TGNN model, a novel training algorithm, and an optimized system.  In 
 experiments, DistTGL achieves near-linear convergence speedup, outperformi
 ng state-of-the-art single-machine method by 14.5% in accuracy and 10.17x 
 in training throughput.\n\nTag: Artificial Intelligence/Machine Learning\n
 \nRegistration Category: Tech Program Reg Pass\n\nReproducibility Badges: 
 Artifact Available\n\nSession Chair: Israt Nisa (Amazon Web Services AI Re
 search and Education)\n\n
END:VEVENT
END:VCALENDAR