Ensembles of randomized trees and their application to image classification
Intervenant(s) : Louis WEHENKEL (University of Liège, Belgium)
The first part of the talk will focus on a recently proposed supervised learning method based on ensembles of totally randomized (regression or classification) trees. The rationale of the method and the geometrical interpretation of its induced models will be discussed and illustrated on simple examples. The second part of the talk will describe a wrapper framework based on this method, devoted to pixel-based image classification. Various real-world applications, extensions, and ongoing work will be briefly mentioned during the presentation.
Javier.Diaz (at) nulllip6.fr