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 @ CUED
SUMMARY:Bayesian nonparametric dynamic-clustering and gene
 tic imputation - Lloyd Elliott (Gatsby Unit\, UCL 
 / Oxford)
DTSTART;TZID=Europe/London:20140304T110000
DTEND;TZID=Europe/London:20140304T120000
UID:TALK50977AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/50977
DESCRIPTION:I will describe new approaches to dynamic-clusteri
 ng based on Bayesian nonparametric (BNP) hidden Ma
 rkov models (HMMs). I will apply these approaches 
 to genotype imputation problems and illustrate the
  practical benefits of BNP. Genetic similarity wit
 hin a population is a function of chromosome posit
 ion and dynamic-clustering based on parametric HMM
 s are popular models of genetic structure. BNP pri
 ors are well suited as extensions of\, or competit
 ors to\, these HMMs because many aspects of geneti
 c processes (such as allele sampling) arise natura
 lly from BNP models. In addition\, BNP priors prov
 ide several practical benefits over parametric HMM
 s. First\, by defining probability distributions o
 n the set of partitions\, BNP priors avoid label s
 witching problems. Second\, costly model selection
  and ad-hoc methods to determine the number of lat
 ent clusters are also avoided. Finally\, the flexi
 bility of BNP often provides state-of-the-art impu
 tation accuracy. I will conclude with directions o
 f future work including the abstraction of the aux
 iliary Gibbs scheme (which I derived for inference
 ) to probabilistic programming for BNP models.
LOCATION:Engineering Department\, CBL Room BE-438
CONTACT:Zoubin Ghahramani
END:VEVENT
END:VCALENDAR