BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Optimal Bayes estimators for block models - Nial Friel (University
  College Dublin)
DTSTART:20160725T150000Z
DTEND:20160725T153000Z
UID:TALK66842@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:In cluster analysis interest lies in probabilistically capturi
 ng partitions of individuals\, items or nodes of a network into groups\, s
 uch that those belonging to the same group share similar attributes or rel
 ational profiles. Bayesian posterior samples for the latent allocation var
 iables can be effectively obtained in a wide range of clustering models\, 
 including finite mixtures\, infinite mixtures\, hidden markov models and b
 lock models for networks. However\, due to the categorical nature of the c
 lustering variables and the lack of scalable algorithms\, summary tools th
 at can interpret such samples are not available. We adopt a Bayesian decis
 ion theoretic approach to define an optimality criterion for clusterings\,
  and propose a fast and context-independent greedy algorithm to find the b
 est allocations. One important facet of our approach is that the optimal n
 umber of groups is automatically selected\, thereby solving the clustering
  and the model-choice problems at the same time. We discuss the choice of 
 loss functions to compare partitions\, and show that our approach can acco
 mmodate a wide range of cases. Finally\, we illustrate our approach on a v
 ariety of real-data applications for the stochastic block model and latent
  block model.  &nbsp\;  <br><br>This is joint work with Riccardo Rastelli.
 &nbsp\;
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR