BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Automatic Differentiation - Part One: A Revisionist History and th
 e State of the Art - Barak A. Pearlmutter\, NUI Maynooth
DTSTART:20140325T100000Z
DTEND:20140325T110000Z
UID:TALK50940@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Automatic Differentiation (aka Algorithmic Differentiation\, a
 ka Computational Differentiation\, aka AD) is an established discipline co
 ncerning methods of transforming algorithmic processes (ie\, computer prog
 rams) which calculate numeric functions to also calculate various derivati
 ves of interest\, and ways of using such methods. We begin with a discussi
 on of the venerable history of the field\, whose roots go back to the dawn
  of the computer age. There are various "modes" of automatic differentiati
 on\, and we will describe forward\, reverse\, and checkpoint-reverse modes
  in detail. We then turn our attention to existing systems\, with a focus 
 on those able to attain high performance\, and the implementation techniqu
 es which allow this. These implementations have a variety of weaknesses an
 d restrictions which\, we will argue\, have impeded the uptake of AD. We c
 lose with a benchmark comparing the fastest current systems with our own r
 esearch prototype compiler based on principles discussed in Part Two. \n(J
 oint work with Jeffrey Mark Siskind.)\n
LOCATION:Small Lecture Theatre\, Microsoft Research Ltd\, 21 Station Road\
 , Cambridge\, CB1 2FB
END:VEVENT
END:VCALENDAR