EVENTO
Discriminant Principal Components
Tipo de evento: Seminário LNCC
Principal Component Analysis (PCA) is one of the most successful approaches to the problem of creating a low dimensional data representation and interpretation. However, since PCA explains the covariance structure of all the data, its most expressive components are not necessarily the most discriminant ones especially when the distributions of each class are not well separated by the mean difference. In this work, instead of sorting the principal components in decreasing order of the corresponding eigenvalues, we propose the idea of using the discriminant weights given by separating hyperplanes to select the most discriminant principal components. To evaluate the discriminant principal components, two classification tasks have been performed using frontal face images: female versus male (gender) experiments, and non-smiling versus smiling (expression) experiments. Our experimental results show that the discriminant principal components have the same low dimensional data representation as the standard principal components, while allowing robust discriminant reconstruction and interpretation of the sample groups and higher recognition rates.
Data Início: 09/07/2007 Hora: 14:00 Data Fim: Hora: 15:30
Local: LNCC - Laboratório Nacional de Computação Ciêntifica - Auditorio A
Comitê Organizador: Carlos Eduardo Thomaz - Centro Universitario da FEI - FEI - cet@fei.edu.br
Apoio Financeiro: PCI
