BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Erin Carson (Charles University)
DTSTART:20240424T070000Z
DTEND:20240424T080000Z
DTSTAMP:20260422T122314Z
UID:MathMAC/11
DESCRIPTION:Title: <a href="https://researchseminars.org/talk/MathMAC/11/"
 >Balancing Inexactness in Matrix Computations</a>\nby Erin Carson (Charles
  University) as part of Modelling of materials - theory\, model reduction 
 and efficient numerical methods (UNCE MathMAC)\n\n\nAbstract\nOn supercomp
 uters that exist today\, achieving even close to the peak performance is i
 ncredibly difficult if not impossible for many applications. Techniques de
 signed to improve the performance of matrix computations- making computati
 ons less expensive by reorganizing an algorithm\, making intentional appro
 ximations\, and using lower precision- all introduce what we can generally
  call “inexactness”. The questions to ask are then:\n1. With all these
  various sources of inexactness involved\, does a given algorithm still ge
 t close enough to the right answer?\n2. Given a user constraint on require
 d accuracy\, how can we best exploit and balance different types of inexac
 tness to improve performance?\nStudying the combination of different sourc
 es of inexactness can thus reveal not only limitations\, but also new oppo
 rtunities for developing algorithms for matrix computations that are both 
 fast and provably accurate. We present few recent results toward this goal
 \, involving mixed precision randomized decompositions and mixed precision
  sparse approximate inverse preconditioners.\n
LOCATION:https://researchseminars.org/talk/MathMAC/11/
END:VEVENT
END:VCALENDAR