Monte Carlo Sampling

[MoumouZhou]

Hello, I would like to ask, does the software support Monte Carlo sampling? I would like to use probability distributions via Monte Carlo sampling as a priori data for multiple input samples of a Bayesian network before performing forward/backward inference (to count the results of the probability distribution of the top event)? If this is possible, could you please let me know if there is any help for this section in the GeNle guidebook, as I would like to go in this direction but am just learning about it and don't really understand it yet. Thanks for your guidance, looking forward to your reply.

[marek [BayesFusion]]

Most certainly. GeNIe and SMILE include several Monte Carlo sampling algorithms and produce samples for each of the nodes in the model, whether the model is discrete, continuous or hybrid. Please look at the section on algorithms for Bayesian networks.
Cheers,

Marek

[MoumouZhou]

Trouble I would like to ask for advice, instead of using a single value for the probability value of the bottom event, I use a probability distribution, how is this probability distribution function constructed? Like if I have a mean value for the failure probability of the base event, and then I solve for the upper and lower values of the failure probability of the base event by fuzzy-fuzzying the failure probability of each base event through a trigonometric fuzzy function, how do I make use of these pre-existing conditions? Is it possible to introduce a parameter setting for the triangular distribution for the bottom events (the probability values of the bottom events are generated based on their probability distribution) and calculate a sample set of failure probabilities for the top events.

[marek [BayesFusion]]

I hope I understand your query correctly, as it contains several terms that could mean various things (like "bottom event", "base event", "upper and lower values", "failure probability", "fuzzy function"). It seems to me that what you want to do cannot be done in GeNIe but you can do almost anything with SMILE. After running a sampling algorithm, you can access raw samples that are stored in each node, which you can intreprete the way you want. GeNIe displays these samples as histograms and they give you an idea of the marginal distributions. Processing, fuzzyfying, truncating, etc., these samples would be something that you need to do yourself. You can also write and use your own inference algorithm while using the network representation offered by SMILE.
I hope this helps,

Marek