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:20260422T000604Z
LOCATION:E Concourse
DTSTART;TZID=America/Denver:20231115T100000
DTEND;TZID=America/Denver:20231115T170000
UID:submissions.supercomputing.org_SC23_sess301_drs123@linklings.com
SUMMARY:I/O Efficient Machine Learning
DESCRIPTION:Meghana Madhyastha (Johns Hopkins University, Argonne National
  Laboratory (ANL))\n\nMy research focuses on systems optimizations for mac
 hine learning, specifically on I/O efficient model storage and retrieval.\
 n\nThe first part of my work focuses on efficient inference serving of tre
 e ensemble models. Tree structures are inherently not cache friendly and t
 heir traversal incurs random I/Os. We developed two systems - Blockset (Bl
 ock Aligned Serialized Trees) and T-REX (Tree Rectangles).\n\nBlockset imp
 roves inference latency in the scenario where the model doesn’t fit in mem
 ory. It introduces the concept of selective access for tree ensembles in w
 hich only the parts of the model needed for inference are deserialized and
  loaded into memory. It uses principles from external memory algorithms to
  rearrange tree nodes in a block aligned format to minimize the number of 
 I/Os needed for inference. T-REX optimizes inference latency for both in-m
 emory inference as well as inference when the model doesn’t fit in memory.
  T-REX reformulates decision tree traversal as hyperrectangle enclosure qu
 eries using the fact that decision trees partition the space into convex h
 yperrectangles.  The test points are then queried for enclosure inside the
  hyperrectangles. In doing random I/O is traded for additional computation
 .\n\nThe second part of my work focuses on efficient deep learning model s
 torage. We implemented a deep learning model repository that requires fine
 -grained access to individual tensors in models. This is useful in applica
 tions such as transfer learning, where individual tensors in layers are tr
 ansferred from one model to another. We’re currently working on caching an
 d prefetching popular tensors based on application level hints.\n\nTag: Ac
 celerators, Artificial Intelligence/Machine Learning, Applications, Cloud 
 Computing, Distributed Computing, Data Analysis, Visualization, and Storag
 e, Data Compression, Heterogeneous Computing, I/O and File Systems, Quantu
 m Computing, Reproducibility, Security, Software Engineering\n\nRegistrati
 on Category: Tech Program Reg Pass, Exhibits Reg Pass\n\n
END:VEVENT
END:VCALENDAR