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:20240626T180033Z
LOCATION:3003\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240626T103000
DTEND;TZID=America/Los_Angeles:20240626T104500
UID:dac_DAC 2024_sess118_RESEARCH1448@linklings.com
SUMMARY:Towards Redundancy-Free Recommendation Model Training via Reusable
 -aware Near-Memory Processing
DESCRIPTION:Research Manuscript\n\nHaifeng Liu, Long Zheng, Yu Huang, Haoy
 an Huang, Xiaofei Liao, and Jin Hai (Huazhong University of Science and Te
 chnology)\n\nThe memory-intensive embedding layer in the recommendation mo
 del continues to be the performance bottleneck. While prior works have att
 empted to improve the embedding layer performance by exploiting the data l
 ocality to cache the frequently accessed embedding vectors and their parti
 al sums. However, these solutions rely on the static cache, which is inval
 idated in the embedding training scenario of the embedding vectors being u
 pdated frequently. To this end, this paper proposes ReFree, a redundancy-f
 ree near-memory processing (NMP) solution for embedding training. Specific
 ally, ReFree identifies the reusable data in real-time for both forward an
 d backpropagation of the embedding layer training, and leverages a lightwe
 ight NMP architecture to enable redundancy-free near-memory acceleration o
 f the entire embedding training process. Evaluation results on real-world 
 datasets show that ReFree outperforms the state-of-the-art solutions by 10
 .9x and reduces 5.3x energy consumption on average.\n\nTopic: Design\n\nKe
 yword: In-memory and Near-memory Computing Architectures, Applications and
  Systems\n\nSession Chairs: Seokhyeong Kang (Pohang University of Science 
 and Technology (POSTECH)) and Giacomo Pedretti (Hewlett Packard Enterprise
 )
END:VEVENT
END:VCALENDAR