BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230831T095746Z
LOCATION:Schwarzhorn
DTSTART;TZID=Europe/Stockholm:20230627T150000
DTEND;TZID=Europe/Stockholm:20230627T153000
UID:submissions.pasc-conference.org_PASC23_sess183_pap114@linklings.com
SUMMARY:Streaming Generalized Canonical Polyadic Tensor Decompositions
DESCRIPTION:Paper\n\nEric Phipps (Sandia National Laboratories), Nicholas 
 Johnson (Cerebras Systems Inc), and Tamara Kolda (MathSci.ai)\n\nIn this p
 aper, we develop a method which we call OnlineGCP for computing the Genera
 lized Canonical Polyadic (GCP) tensor decomposition of streaming data. GCP
  differs from traditional canonical polyadic (CP) tensor decompositions as
  it allows for arbitrary objective functions which the CP model attempts t
 o minimize. This approach can provide better fits and more interpretable m
 odels when the observed tensor data is strongly non-Gaussian. In the strea
 ming case, tensor data is gradually observed over time and the algorithm m
 ust incrementally update a GCP factorization with limited access to prior 
 data. In this work, we extend the GCP formalism to the streaming context b
 y deriving a GCP optimization problem to be solved as new tensor data is o
 bserved, formulate a tunable history term to balance reconstruction of rec
 ently observed data with data observed in the past, develop a scalable sol
 ution strategy based on segregated solves using stochastic gradient descen
 t methods, describe a software implementation that provides performance an
 d portability to contemporary CPU and GPU architectures and demonstrate th
 e utility and performance of the approach and software on several syntheti
 c and real tensor data sets.\n\nDomain: Engineering, Computer Science, Mac
 hine Learning, and Applied Mathematics 
\n\nSession Chair: Johannes Gebert
  (High-Performance Computing Center Stuttgart, University of Stuttgart)
END:VEVENT
END:VCALENDAR