par Hsia, Yen-Teh 
Référence Lecture notes in computer science, 521 LNCS, page (68-74)
Publication Publié, 1991

Référence Lecture notes in computer science, 521 LNCS, page (68-74)
Publication Publié, 1991
Article révisé par les pairs
| Résumé : | Dempster’s rule of combination, the main inference mechanism of the Dempster-Shafer theory of belief functions [Shafer 76; Smets 88], requires that the belief functions to be combined must be “independent”. This independence assumption is usually understood to be composed of two parts: (1) the uniqueness assumption, which states that each belief function to be combined is based on a unique body of evidence, and (2) the evidential independence (or distinctness) assumption, which states that we must be able to argue about the independence of all these different bodies of evidence. In this paper, we suggest that the uniqueness assumption is not necessary, and that it should be alright if we can come up with “independent specifications” of beliefs. |



