Embedding sample points uncertainty measures in learning algorithms

Learning algorithms consider a sample consisting of pairs (pattern, label) and output a decision rule, possibly: (i) associating each pattern with the corresponding label, and (ii) generalizing to new patterns drawn from the same distribution of the original sample. This work proposes a set of methodologies to be applied to existing learning strategies in order to deal with more complex kinds of data sets, carrying also a quantitative measure on the quality of each label.