BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Automating stochastic gradient methods with adaptive batch sizes -
  Tom Goldstein (University of Maryland\, College Park)
DTSTART:20170906T085000Z
DTEND:20170906T094000Z
UID:TALK78071@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:This talk will address several issues related to training neur
 al networks using stochastic gradient methods.&nbsp\; First\, we&#39\;ll t
 alk about the difficulties of training in a distributed environment\, and 
 present a new method called centralVR for boosting the scalability of trai
 ning methods.&nbsp\; Then\, we&#39\;ll&nbsp\;talk about the issue of autom
 ating stochastic gradient descent\, and show that learning rate selection 
 can be simplified using "Big Batch" strategies that adaptively choose mini
 batch sizes.<br>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR