Fusion, propagation, and structuring in belief networks
Artificial Intelligence
Logic for problem-solving
Programming in Prolog
Probabilistic reasoning in expert systems: theory and algorithms
Probabilistic reasoning in expert systems: theory and algorithms
Probabilistic Horn abduction and Bayesian networks
Artificial Intelligence
Fril- Fuzzy and Evidential Reasoning in Artificial Intelligence
Fril- Fuzzy and Evidential Reasoning in Artificial Intelligence
Qualitative relevance and independence: a roadmap
IJCAI'97 Proceedings of the 15th international joint conference on Artifical intelligence - Volume 1
Hi-index | 0.00 |
Causal nets (Pearl 1986) are an elegant way of representing the structure and relationships of a set of data. The propagation of changes through the net has been examined and reported on in many works (Pearl 1986, Lauritzen & Speigelhalter 1988, Neapolitan 1990). Causal nets are defined by the properties of conditional independence, and so the structure of the net may be obtained by discovering conditional independences. Many of the examples in the literature test for complete equality. However, the presence of noise and the unreliability of comparing two real numbers means that equality is taken to mean equality within a particular tolerance. Where a set of data contains a number of representative subsets this tolerance can be almost zero. If there is an incomplete subset in the data and conjoint events then the tolerance cannot be zero. The paper presents a method for estimating the size of the partial cohort, the size of the representative cohorts and thus provides a robust test for conditional independence.