Choosing Abstractions for Hierarchical Diagnosis
Résumé
This paper deals with the choice of abstractions for stating hierarchical diagnosis problems. Generally, hierarchical models are built manually and choosing the appropriate abstractions is quite an empirical science. To tackle this issue, we frame a diagnosis problem as an optimal constraint satisfaction problem (OCSP) and we define abstraction related to two OCSP's, in the structure and in the search space. This allow us to analyse the influence of the abstraction on the temporal computational complexity reduction offered by hierarchical reasoning. Optimal abstractions are shown to be built on the well-known diagnosis concept of potential conflict.
Domaines
Intelligence artificielle [cs.AI]Origine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...