Properties of separation procedures for discrete objects in Bayesian network models
- 46 Downloads
It is shown that for discrete objects constructed on bounded samples there are examples that support the faithfulness assumption and examples for which it fails. Thus, the properties of separation procedures for continuous models do not hold for discrete objects.
Keywordsseparation Bayesian network Markov condition discrete object
Unable to display preview. Download preview PDF.
- 1.J. Pearl, Causality: Models, Reasoning, and Inference, Cambridge Univ. Press (2000).Google Scholar
- 3.R. E. Neapolitan, Learning Bayesian Networks, Prentice Hall, NJ, Upper Saddle River (2004).Google Scholar
- 6.V. Boss, Lectures in Mathematics, Vol. 12, Counterexamples and Paradoxes. A Handbook, Knizhnyi Dom Librokom, Moscow (2009).Google Scholar