BEGIN:VCALENDAR
PRODID:-//Silva News Calendaring//lonely event//
VERSION:2.0
BEGIN:VEVENT
CREATED:20111013T092549Z
DESCRIPTION:<br /><br /><strong>Orador Convidado</strong>: Carlo Gatta (In
 vestigador do Centro de Visão Computacional - CVC - da Universidade de Ba
 rcelona)<br /><br /><strong>Titulo</strong>: "Multiscale Stacked\nSequenti
 al Learning"<br /><br /><strong>Abstract</strong>: The classification and/
 or segmentation of\nsequential data (such as text\, audio\, images and vid
 eos) can be improved by\nmoving from i.i.d. classifiers to contextual-awar
 e methods. One popular\napproach to contextual-aware classification is the
  Conditional Random Fields\n(CRF)\, or other related graph-based algorithm
 s. In this seminar we will show a\nmethod\, called “Multi-scale Stacked 
 Sequential Learning” (MSSL)\, which allows\nconsidering the context in a
  very efficient way. The method proved to\noutperform CRFs in several task
 s\, in terms of accuracy\, precision\, sensitivity.\nTraining and testing 
 times are more than one order smaller than CRFs. We tested\nthe method on 
 text classification\, image segmentation (binary and 8 classes\nproblems)\
 , temporal classification\, and volumetric classification. Moreover\,\nits
  implementation is straightforward (less than half an our in MATLAB and\n&
 lt\;100 lines of code) and\, being a meta-learning scheme\, any classifier
  can be\nemployed in its use (kNN\, Adaboost\, SVMs\, etc…). The meta-le
 arner MATLAB code\nwill be provided to the audience.
DTEND:20111014T110000Z
DTSTART:20111014T100000Z
LAST-MODIFIED:20111013T093509Z
LOCATION:FCTUC - DEI\, Sala E4.7
SUMMARY:Palestra Convidada - "Multiscale Stacked Sequential Learning"
UID:1851572959@0@silvanews
URL:http://www.uc.pt/fctuc/dei/noticias_dei/palestraconvidada6
END:VEVENT
END:VCALENDAR