Supervised learning in presence of outliers, label noise and unobserved classes

Three important issues are often encountered in Supervised Classification: class-memberships are unreliable for some training units (Label Noise), a proportion of observations might depart from the bulk of the data structure (Outliers) and groups represented in the test set may have not been encountered earlier in the learning phase (Unobserved Classes). The present work introduces a Robust and AdaptiveEigenvalue-Decomposition Discriminant Analysis (RAEDDA) capable of handling situations in which one or more of the afore-described problems occur. Transductiveand inductive robust EM-based procedures are proposed for parameter estimation and experiments on real data, artificially adulterated, are provided to underline the benefits of the proposed method.