BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium: OpAMP: Approximate Message Passing with Time-Varyin
 g Linear Operators
DTSTART:20261001T161500
DTEND:20261001T171500
DTSTAMP:20261023T130140Z
UID:d2063c36aebccdacdef155937eccadc04b660436da23d0f388666f59
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: Bobak Nazer - Boston University\nVideo of his talk\n\nAb
 stract\nApproximate message passing (AMP) is a powerful theoretical framew
 ork that provides precise characterizations of iterative inference algorit
 hms in the high-dimensional regime\, including generalized linear regressi
 on and low-rank matrix estimation as special cases. Implicitly\, the AMP f
 ramework presumes that computations (e.g.\, matrix multiplications) are ca
 rried out in a centralized fashion. However\, in practice\, due to time or
  memory constraints\, large-scale computations are often distributed acros
 s multiple servers\, some of which may respond more slowly than others (i.
 e.\, stragglers). We demonstrate how to generalize AMP to allow for a time
 -varying linear operator to be applied to the data matrix at each iteratio
 n\, capturing erasures due to missing partial computations as a special ca
 se. One surprising consequence is that\, in certain applications\, schedul
 ing partial computations can be more efficient in terms of total matrix mu
 ltiplications. We illustrate this phenomenon via a case study of power ite
 ration applied to a spiked Wigner model. Joint work with Riccardo Rossetti
  and Galen Reeves.\n\nBio\nBobak Nazer is an Associate Professor in the De
 partment of Electrical and Computer Engineering at Boston University\, and
  is spending the 2026 - 2027 academic year on sabbatical at EPFL. He recei
 ved the Ph.D. degree in 2009 and the M.S. degree in 2005 from the Universi
 ty of California\, Berkeley\, as well as the B.S. degree in 2003 from Rice
  University\, all in electrical engineering. He is a past recipient of the
  IEEE Communications Society & Information Theory Society Joint Paper Awar
 d\, the NSF CAREER Award\, the Eli Jury Award\, and the Gerald and Deanne 
 Gitner Family Award for Innovation in Teaching with Technology.\n\nMore in
 formation
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420 https://epfl.zoom.us/
 j/66735896236
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR