Classifier Selection
Ludmila I. Kuncheva · 2014
The presumption in classifier selection is that there is an oracle that can identify the best expert for a particular input x. This expert's decision is accepted as the decision of the ensemble for x. This chapter address the following questions: (1) how do we build the individual classifiers; (2) should they be stable or unstable; (3) homogeneous or heterogeneous; (4) how do we evaluate the competence of the classifiers for a given x; and (5) once the competences are found, what selection strategy shall we use. This chapter shows the operation of a classifier selection, and demonstrates the rationale for classifier selection ensembles. It also discusses an interesting ensemble method which belongs to the classifier selection group — the so-called mixture of experts (ME), and the cascade classifiers, which can be thought of as a version of a classifier selection ensemble.