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:Machine Learning Reading Group @ CUED
SUMMARY:Stein Discrepancy - Sebastian Ober (University of
Cambridge)
DTSTART;TZID=Europe/London:20190128T140000
DTEND;TZID=Europe/London:20190128T153000
UID:TALK119455AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/119455
DESCRIPTION:Recently\, Stein discrepancy-based methods have be
come popular tools for machine learning applicatio
ns\, including verifying the convergence of MCMC [
1]\, goodness-of-fit tests [2][3]\, and variationa
l inference [4][5]. Although Stein's method has be
en long-known\, widespread use was limited by an i
ntractable optimization over a difficult function
space. However\, the recent development of the ke
rnelized Stein discrepancy (KSD) [2][3] has circum
vented this difficulty. Our talk will give a theor
etical introduction to the Stein discrepancy and K
SD. We will then introduce two recent applications
of the Stein discrepancy to machine learning prob
lems. The first of these\, Stein variational gradi
ent descent (SVGD) [4]\, shows how to apply the KS
D to variational inference.\nWe conclude by discus
sing the Stein variational autoencoder (Stein VAE)
[5]\, which applies SVGD to VAE learning. \n\nPap
ers that are important for the talk:\n\n\n[2] Chwi
alkowski\, K.\, Strathmann\, H.\, and Gretton\, A.
A kernel test of goodness of fit. In ICML\, 2016.
https://arxiv.org/abs/1602.02964\n\n[4] Liu\, Q.
and Wang\, D. Stein variational gradient descent:
A general purpose Bayesian inference algorithm. In
NIPS\, 2016.\nhttps://arxiv.org/abs/1608.04471\n\
n\nAdditional Recommended reading:\n\n[1] Gorham\,
J. and Mackey\, L. Measuring sample quality with
Stein's method. In NIPS\, pp. 226-234\, 2015. http
s://arxiv.org/abs/1506.03039\n-- Note: we will onl
y be discussing up to and including section 3\n\n[
3] Liu\, Q.\, Lee\, J.\, and Jordan\, M. I. A kern
elized Stein discrepancy for goodness-of-fit tests
. In ICML\, 2016.\nhttps://arxiv.org/abs/1602.0325
3\n-- Note: this is essentially the same paper as
[2]\, which is what we will present\n\n\n[5] Pu\,
Y.\, Gan\, Z.\, Henao\, R.\, Li\, C.\, Han\, S.\,
and Carin\, L. VAE learning with Stein variational
gradient descent. In NIPS\, 2017. \nhttps://arxiv
.org/abs/1704.05155
LOCATION:Engineering Department\, CBL Room 438
CONTACT:
END:VEVENT
END:VCALENDAR