BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Accelerated Consensus via Min-Sum Splitting - Patrick Rebeschini (
 Oxford)
DTSTART:20171110T160000Z
DTEND:20171110T170000Z
UID:TALK87061@talks.cam.ac.uk
CONTACT:Quentin Berthet
DESCRIPTION: We apply the Min-Sum message-passing protocol to solve the co
 nsensus problem in distributed optimization. We show that while the ordina
 ry Min-Sum algorithm does not converge\, a modified version of it known as
  Splitting yields convergence to the problem solution. We prove that a pro
 per choice of the tuning parameters allows Min-Sum Splitting to yield subd
 iffusive accelerated convergence rates\, matching the rates obtained by sh
 ift-register methods. The acceleration scheme embodied by Min-Sum Splittin
 g for the consensus problem bears similarities with lifted Markov chains t
 echniques and with multi-step first order methods in convex optimization. 
 (Joint work with Sekhar Tatikonda\, based on arxiv.org/abs/1706.03807\, at
  NIPS 2017)
LOCATION:MR12
END:VEVENT
END:VCALENDAR