ISSN:
1573-0670
Schlagwort(e):
Bayesian inference
;
uncertainty management
;
probabilistic networks
;
relative entropy
Quelle:
Springer Online Journal Archives 1860-2000
Thema:
Informatik
Notizen:
Abstract Uncertainty management is critical to the effective use of knowledge-based systems in a wide variety of domains. Design is typical of these domains in that the implementation of a design in an artifact, the future environment for the artifact, and the component characteristics of the artifact are all uncertain. Existing probabilistic schemes to address the inherent uncertainty in areas like design assume precise knowledge of the probabilities of relevant events. This paper defines a probabilistic method for uncertainty management with imprecise inputs. The approach combines Bayesian inference networks and information theoretic inference procedures. The resulting scheme manages both imprecision and uncertainty in the problem domain. An application of the approach to materiel design is described.
Materialart:
Digitale Medien
URL:
http://dx.doi.org/10.1007/BF00245461
Permalink
