Combining binary classifiers with imprecise probabilities

Sébastien Destercke (1) , Benjamin Quost (1)

1 Heudiasyc - Heuristique et Diagnostic des Systèmes Complexes [Compiègne]

This paper proposes a simple framework to combine binary classifiers whose outputs are imprecise probabilities (or are transformed into some imprecise probabilities, e.g., by using confidence intervals). This combination comes down to solve linear programs describing constraints over events (here, subsets of classes). The number of constraints grows linearly with the number of classifiers, making the proposed framework tractable even for problems involving a relatively large number of classes.

Mots clés: Classification; combinaison; binaire; probabilités imprécises

Domaines: Apprentissage [cs.LG]; Intelligence artificielle [cs.AI]; Autres [stat.ML]

Sébastien Destercke, Benjamin Quost. Combining binary classifiers with imprecise probabilities. Integrated Uncertainty in Knowledge Modelling and Decision Making, Oct 2011, Hangzhou, China. pp.219-230, ⟨10.1007/978-3-642-24918-1_24⟩. ⟨hal-00655600⟩