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:Isaac Newton Institute Seminar Series
SUMMARY:Efficient Bayesian inference of multi-scale networ
k structures - Tiago Peixoto (ISI Foundation\; Uni
versität Bremen )
DTSTART;TZID=Europe/London:20160714T163000
DTEND;TZID=Europe/London:20160714T170000
UID:TALK66760AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/66760
DESCRIPTION:A principled approach to characterize the hidden s
tructure of networks is to formulate generative mo
dels\, and then infer their parameters from data.
When the desired structure is composed of modules
or "communities"\, a popular choice for this task
is the stochastic block model\, where nodes are di
vided into groups\, and the placement of edges is
conditioned on the group memberships. In this talk
\, we will present a nonparametric Bayesian infere
nce framework based on a *microcanonical* for
mulation of the stochastic block model. We show ho
w this simple model variation allows simultaneousl
y for two important improvements over more traditi
onal inference approaches: 1. Deeper Bayesian hier
archies\, with noninformative priors replaced by s
equences of priors and hyperpriors\, that not only
remove limitations that seriously degrade the inf
erence of large networks\, but also reveal structu
res at multiple scales\; 2. A very efficient infer
ence algorithm that scales well not only for netwo
rks with a large number of nodes and edges\, but a
lso with an *unlimited* *number of groups. We show also how this approach can be used to
sample group hierarchies from the posterior distri
bution\, perform model selection\, and how it can
be easily be generalized to networks with edge cov
ariates and \; node annotations. *

LOCATION:Seminar Room 1\, Newton Institute
CONTACT:info@newton.ac.uk
END:VEVENT
END:VCALENDAR