Elicitation of probabilities for belief networks: combining qualitative and quantitative information

摘要: Although the usefulness of belief networks for reasoning under uncertainty is widely accepted, obtaining numerical probabilities that they require still perceived a major obstacle. Often not enough statistical data available to allow reliable probability estimation. Available information may be directly amenable encoding in network. Finally, domain experts reluctant provide probabilities. In this paper, we propose method elicitation from expert non-invasive and accommodates whatever probabilistic willing state. We express all information, whether qualitative or quantitative nature, canonical form consisting (in) equalities expressing constraints on hyperspace possible joint distributions. then use derive second-order distributions over desired