Construction of classifier ensembles

Our contribution to this topic is an overview and an experimental comparison of methods for constructing classifier ensembles, when the combining rule is chosen beforehand (usually, a fixed rule). We focused in particular on the so-called overproduce and choose strategy, which consists first in constructing a large ensemble of classifiers (several methods can be used to this aim) and then in selecting an effective subset of them.

People working on this topic:

• Giorgio Fumera
• Giorgio Giacinto
• Fabio Roli