Thesiss

We study diversity in classifier ensembles from a broader perspectivethan the 0/1 loss function, the main reason being that the bias-variance decomposition of the 0/1 loss function is not unique, and therefore the relationship between ensemble accuracy and diversity is still unc…