BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Information Theory Seminar
SUMMARY:Bayes-optimal estimation in generalized linear mod
els - Ramji Venkataramanan\, Department of Enginee
ring\, Cambridge
DTSTART;TZID=Europe/London:20230215T140000
DTEND;TZID=Europe/London:20230215T150000
UID:TALK197254AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/197254
DESCRIPTION:We consider the problem of signal estimation in g
eneralized linear models (GLM)\, a class of models
which includes canonical problems such as linear
regression\, logistic regression\, and phase retri
eval. Recent work has precisely characterized the
asymptotic minimum mean-squared error (MMSE) for
GLMs with i.i.d. Gaussian measurement matrices.
However\, in many models there is a significant ga
p between the MMSE and the performance of the best
known feasible estimators. To address this\, we c
onsider GLMs defined via _spatially coupled_ measu
rement matrices. We propose an efficient approxima
te message passing (AMP) algorithm for estimation
and prove that the error of a carefully tuned AMP
estimator approaches the asymptotic MMSE. \n\nThe
talk will not assume any background on message pas
sing or spatial coupling. Joint work with Pablo Pa
scual Cobo and Kuan Hsieh.
LOCATION:MR5\, CMS Pavilion A
CONTACT:Prof. Ramji Venkataramanan
END:VEVENT
END:VCALENDAR