Experimental analysis of large belief networks for medical diagnosis
From Semantic Portal Wiki
{{#vardefine:category|Publication}}{{#vardefine:templatename|i.publication}}{{#vardefine:package|smwbp_instance_templates}}
| Edit |
Reference: {{#vardefine:pagename|experimental analysis of large belief networks for medical diagnosis }}
- [[]]
bibtex
{{#vardefine:pagename|Experimental analysis of large belief networks for medical diagnosis }}{{#vardefine:key| }}
abstract: We present an experimental analysis of two parameters that are important inknowledge engineering for large belief networks. We conducted the experimentson a network derived from the Internist-1 medical knowledge base. In thisnetwork, a generalization of the noisy-OR gate is used to model causalindependence for the multi-valued variables, and leak probabilities are used torepresent the nonspecified causes of intermediate states and findings. We studytwo network parameters, (1) the parameter governing the assignment ofprobability values to the network, and (2) the parameter denoting whether thenetwork nodes represent variables with two or more than two values. Theexperimental results demonstrate that the binary simplification computesdiagnoses with similar accuracy to the full multivalued network. We discuss theimplications of these parameters, as well as other network parameters, forknowledge engineering for medical applications.
download:
- paper:
- slides:
| Abstract | We present an experimental analysis of two … We present an experimental analysis of two parameters that are important inknowledge engineering for large belief networks. We conducted the experimentson a network derived from the Internist-1 medical knowledge base. In thisnetwork, a generalization of the noisy-OR gate is used to model causalindependence for the multi-valued variables, and leak probabilities are used torepresent the nonspecified causes of intermediate states and findings. We studytwo network parameters, (1) the parameter governing the assignment ofprobability values to the network, and (2) the parameter denoting whether thenetwork nodes represent variables with two or more than two values. Theexperimental results demonstrate that the binary simplification computesdiagnoses with similar accuracy to the full multivalued network. We discuss theimplications of these parameters, as well as other network parameters, forknowledge engineering for medical applications. edge engineering for medical applications. |
| Author | Malcolm Pradhan +, Gregory M. Provan +, and Max Henrion + |
| Bibtype | techreport + |
| Institution | Knowledge Systems, AI Laboratory + |
| Key | KSL-94-36 + |
| Month | May + |
| Note | Medical Computer Science + |
| Number | KSL-94-36 + |
| Tag | Computer science + |
| Title | Experimental Analysis of Large Belief Networks for Medical Diagnosis + |
| Tr id | KSL-94-36 + |
| Year | 1994 + |
