This work deals with a robustification of classification analysis for data with a high dimension. We combine a robust principal component analysis and robust classification.
This method inherits the idea of down-weighting less reliable data points from the least weighted squares regression.