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:CCIMI Short Course: Tamara Broderick (MIT)
SUMMARY:Variational Bayes and Beyond: Foundations of Scala
ble Bayesian Inference - Tamara Broderick (MIT)
DTSTART;TZID=Europe/London:20200115T140000
DTEND;TZID=Europe/London:20200115T160000
UID:TALK136285AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/136285
DESCRIPTION:Bayesian methods exhibit a number of desirable pro
perties\nfor modern data analysis---including (1)
coherent quantification of uncertainty\, (2) a mod
ular modeling framework able to capture complex ph
enomena\, (3) the ability to incorporate prior inf
ormation from an expert source\, and (4) interpret
ability. In practice\, though\, Bayesian\ninferenc
e necessitates approximation of a high-dimensional
integral\, and some traditional algorithms for th
is purpose can be slow---notably at data scales of
current interest. The tutorial will cover the fou
ndations of some modern tools for fast\, approxima
te Bayesian inference at scale. One increasingly p
opular framework is provided by\n"variational Baye
s" (VB)\, which formulates Bayesian inference as a
n optimization problem. We will examine key benefi
ts and pitfalls of using VB in practice\, with a f
ocus on the widespread "mean-field variational Bay
es" (MFVB) subtype. We will highlight properties t
hat anyone working with VB\, from the data analyst
to the theoretician\, should be aware of. And we
will discuss a number of open challenges.
LOCATION:MR15 (GL.02)\, Pavilion G\, CMS
CONTACT:J.W.Stevens
END:VEVENT
END:VCALENDAR