Linear methods of dimension reduction for classification

For classification problems, traditional dimension reduction methods often take into account only the feature information, while ignoring the class label. This poses an opportunity for improvement. In this thesis, we explore new methods that aim to find linear orthogonal projections that maximize op...