Seminář: Robust Classifiers in Multivariate Statistics and Machine Learning

Various methods multivariate statistics and data mining suffer from the presence of outlying measurements (outliers) in the data. Therefore, attention has been paid to proposing such robust alternatives which are resistant (insensitive) to data contamination. Two particular methods will be presented in the talk, namely (1) a robust version of support vector machine classifier, and (2) our proposal of a robust linear discriminant analysis. The latter method is also suitable for high-dimensional data with the number of variables exceeding the number of cases (observations). The talk is focused on the methodology and principles, but illustrative examples will be also presented.