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:20200129T163227Z
LOCATION:301-302-303
DTSTART;TZID=America/Denver:20191120T140000
DTEND;TZID=America/Denver:20191120T143000
UID:submissions.supercomputing.org_SC19_sess142_pap513@linklings.com
SUMMARY:An Efficient Mixed-Mode Representation of Sparse Tensors
DESCRIPTION:Paper\n\nAn Efficient Mixed-Mode Representation of Sparse Tens
ors\n\nNisa, Li, Sukumaran-Rajam, Rawat, Krishnamoorthy...\n\nThe Compress
ed Sparse Fiber (CSF) representation for sparse tensors is a generalizatio
n of the Compressed Sparse Row (CSR) format for sparse matrices. For a te
nsor with d modes, typical tensor methods such as CANDECOMP/PARAFAC decomp
osition (CPD) require a sequence of d tensor computations, where efficient
memory access with respect to different modes is required for each of the
m. The straightforward solution is to use d distinct representations of t
he tensor, with each one being efficient for one of the d computations. H
owever, a d-fold space overhead is often unacceptable in practice, especia
lly with memory-constrained GPUs. \n\nIn this paper, we present a mixed-mo
de tensor representation that partitions the tensor's nonzero elements int
o disjoint sections, each of which is compressed to create fibers along a
different mode. Experimental results demonstrate that better performance
can be achieved while utilizing only a small fraction of the space require
d to keep d distinct CSF representations.\n\nTag: Tech Program Reg Pass, A
lgorithms, Linear Algebra, Performance, Sparse Computation\n\nRegistration
Category: Tech Program Reg Pass, Algorithms, Linear Algebra, Performance,
Sparse Computation
URL:https://sc19.supercomputing.org/presentation/?id=pap513&sess=sess142
END:VEVENT
END:VCALENDAR