Apprentissage des paramètres des règles d’un système à base de règles possibilistes

In this paper, we introduce a paradigm for learning the parameters of the rules of a possibilistic rule-based system according to a training datum. For any system composed of n parallel possibilistic rules, we introduce a system of equations denoted (?n), analogous to the Farreny- Prade equation system, where the unknown is a vector whose components are the parameters of the rules that must be determined according to a training datum. We establish the necessary and sufficient conditions for the system (?n) to have solutions. In this case, we show that there is a unique maximal and a unique minimal solution. Finally, our results are illustrated by an example.