Contribution to Feature and Instance Selection for Visual Data Analysis
Fadi Dornaika, Department of Computer Science & Artificial Intelligence, University of the Basque Country, Espagne
mardi 29 mars 2016 à 09h45
Feature selection and instance selection are two important topics in the domains of data mining and pattern recognition. In this talk, I will briefly present some contribution to feature selection in embedded spaces provided by manifold learning paradigms. More precisely, I will present selection schemes that can work for a two-stage embedding scheme such as PCA+LDA. Then, I will present a cooperative feature selection-manifold learning framework. Finally, I will present two different methods for instance selection that are based on data self-representativeness adopting a block sparsity constraint. The talk shows some experimental results in tackling the problems of face recognition and image classification.