On the Spectral Decomposition in Normal Discriminant Analysis

• DOI: 10.1080/03610918.2012.735318
• Publication Date: 2013-07-25T17:46:51Z
• Authors: Luca Bagnato; Francesca Greselin; Antonio Punzo
• Keywords: Optimal discriminant analysis

Abstract This article enlarges the covariance configurations, on which classical linear discriminant analysis is based, by considering four models arising from spectral decomposition when eigenvalues and/or eigenvectors matrices are allowed to vary or not between groups. As in approach, assessment of these configurations accomplished via a test training set. The discrimination rule then built upon configuration provided test, unlabeled data. Numerical experiments, simulated and real data, have been performed evaluate gain our proposal with respect analysis. Keywords: CEM algorithmEM algorithmMixture modelsMultiple testing proceduresNormal analysisSpectral decompositionMathematics Subject Classification: 62F0362F0762F3062H30