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:Optimal Bayes estimators for block models - Nial F
riel (University College Dublin)
DTSTART;TZID=Europe/London:20160725T160000
DTEND;TZID=Europe/London:20160725T163000
UID:TALK66842AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/66842
DESCRIPTION:In cluster analysis interest lies in probabilistic
ally capturing partitions of individuals\, items o
r nodes of a network into groups\, such that those
belonging to the same group share similar attribu
tes or relational profiles. Bayesian posterior sam
ples for the latent allocation variables can be ef
fectively obtained in a wide range of clustering m
odels\, including finite mixtures\, infinite mixtu
res\, hidden markov models and block models for ne
tworks. However\, due to the categorical nature of
the clustering variables and the lack of scalable
algorithms\, summary tools that can interpret suc
h samples are not available. We adopt a Bayesian d
ecision theoretic approach to define an optimality
criterion for clusterings\, and propose a fast an
d context-independent greedy algorithm to find the
best allocations. One important facet of our appr
oach is that the optimal number of groups is autom
atically selected\, thereby solving the clustering
and the model-choice problems at the same time. W
e discuss the choice of loss functions to compare
partitions\, and show that our approach can accomm
odate a wide range of cases. Finally\, we illustra
te our approach on a variety of real-data applicat
ions for the stochastic block model and latent blo
ck model. \;

This is joint work wit
h Riccardo Rastelli. \;
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:INI IT
END:VEVENT
END:VCALENDAR