I had recently asked this question where problems encoding a Bayesian Network were linked to the use of MultinomialDistribution. While that problem can be avoided using EmpiricalDistribution, there remains an issue with using ProbabilityDistribution for larger networks as it seems: While Probability can be used for inference with 4 nodes, it will not evaluate for the "full" example network of 5 nodes -- which still is far removed from real application demands. Why is this so? What can be done about it?

I had recently asked this question where problems encoding a Bayesian Network were linked to the use of MultinomialDistribution. While that problem can be avoided using EmpiricalDistribution, there remains an issue with using ProbabilityDistribution for larger networks as it seems: While Probability can be used for inference with 4 nodes, it will not evaluate for the "full" example network of 5 nodes -- which still is far removed from real application demands. Why is this so? What can be done about it?