Many present day applications of statistical learning involve large numbers of predictor variables. Often, that number is much larger than the number of cases or observations available for training the learning algorithm. In such situations, traditional methods fail. Recently, new techniques...