BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:From Automated Currency Validation to Protein Fold Recognition: Pr
 obabilistic Multi-class Multi-kernel Learning - Theodoros Damoulas (CS\, U
 niversity of Glasgow)
DTSTART:20080730T130000Z
DTEND:20080730T140000Z
UID:TALK12651@talks.cam.ac.uk
CONTACT:David MacKay
DESCRIPTION:In diverse machine learning problems ranging from automated cu
 rrency validation (ACV) to protein fold prediction\, we encounter the situ
 ation where multiple object descriptors are available for a possibly multi
 nomial classification task. Specifically\, ACV considers the challenging
  and unresolved problem of counterfeit note detection while depositing cur
 rency in an ATM that is equipped with a plurality of sensors. In an analog
 ous manner\, when predicting the structural fold of a protein multiple fea
 ture sets are available\, ranging from global characteristics like the ami
 no-acid composition and predicted secondary structure\, to attributes deri
 ved from local sequence alignment such as the Smith-Waterman scores. \nThe
 se problems raise the need for a classification method that is able to as
 sess the contribution of these potentially heterogeneous object descriptor
 s while utilizing such information to improve predictive performance. \nIn
  this talk I will present a hierarchical Bayesian multi-class multi-kernel
  pattern recognition machine that informatively combines the available fea
 ture groups and\, as is demonstrated\, is able to provide the state-of-the
 -art in performance accuracy on the problems considered. The full Markov c
 hain Monte Carlo solution of the model is offered via a Metropolis-Hastin
 gs within Gibbs sampling \nprocedure and also an efficient variational Bay
 es approximation is proposed. \n\n
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR