B1381
Title: Robust envelope discriminant analysis
Authors: Abdul-Nasah Soale - Case Western Reserve University (United States) [presenting]
Abstract: Classical linear discriminant analysis imposes an assumption of normality on the conditional distribution of the predictors given the classes. However, in practice, this assumption is easily violated. Motivated by the recent work of envelope discriminant analysis, our robust envelope proposal extends the estimation of the envelope discriminant subspace beyond the normality assumption. We demonstrate the promising performance of our proposal using both synthetic and real data analysis.
