BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Suvrit Sra (MIT)
DTSTART:20210303T204500Z
DTEND:20210303T214500Z
DTSTAMP:20260414T173728Z
UID:BUGeom/18
DESCRIPTION:Title: <a href="https://researchseminars.org/talk/BUGeom/18/">
 Accelerated gradient methods on Riemannian manifolds</a>\nby Suvrit Sra (M
 IT) as part of Boston University Geometry/Physics Seminar\n\nLecture held 
 in Zoom meeting ID: 974 5641 9902.\n\nAbstract\nThis talk lies at the inte
 rface of geometry and optimization. I'll talk about geodesically convex op
 timization problems\, a rich class of non-convex optimization problems tha
 t admit tractable global optimization. I'll provide some background on thi
 s class and some motivating examples. Beyond a general introduction to the
  topic area\, I will dive deeper into a recent discovery of a long-sought 
 result: an accelerated gradient method for Riemannian manifolds. Towards d
 eveloping this method\, we will revisit Nesterov's (Euclidean) estimate se
 quence technique and present a conceptually simple alternative. We will th
 en generalize this simpler alternative to the Riemannian setting. Combined
  with a new geometric inequality\, we will then obtain the first (global) 
 accelerated Riemannian-gradient method. I'll also comment on some very rec
 ent updates on this topic.\n
LOCATION:https://researchseminars.org/talk/BUGeom/18/
END:VEVENT
END:VCALENDAR