BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160050Z
LOCATION:Track 4
DTSTART;TZID=America/New_York:20201117T140000
DTEND;TZID=America/New_York:20201117T143000
UID:submissions.supercomputing.org_SC20_sess176_pap223@linklings.com
SUMMARY:Kraken: Memory-Efficient Continual Learning for Large-Scale Real-T
 ime Recommendation
DESCRIPTION:Paper\n\nKraken: Memory-Efficient Continual Learning for Large
 -Scale Real-Time Recommendation\n\nXie, Ren, Lu, Yang, Xu...\n\nModern rec
 ommendation systems in industry often use deep learning (DL) models that a
 chieve better model accuracy with more data and model parameters. Current 
 open-source DL frameworks, however, such as TensorFlow and PyTorch, show r
 elatively low scalability on training recommendation models with terabytes
  of parameters. To efficiently learn large-scale recommendation mode
 ls from data streams that generate hundreds of terabytes training data dai
 ly, we introduce a continual learning system called Kraken. Kraken contain
 s a special parameter server implementation that dynamically adapts to the
  rapidly changing set of sparse features for the continuous training and s
 erving of recommendation models. Kraken provides a sparsity-aware training
  system that uses different learning optimizers for dense and sparse param
 eters to reduce memory overhead. Extensive experiments using real-world da
 tasets confirm the effectiveness and scalability of Kraken. Kraken c
 an benefit the accuracy of recommendation tasks with the same memory
  resources, or trisect the memory usage, while keeping model performance.\
 n\nTag: Machine Learning, Deep Learning and Artificial Intelligence, Memor
 y Optimization, Scalable Computing\n\nRegistration Category: Tech Program 
 Reg Pass
END:VEVENT
END:VCALENDAR