calculating inferences

[megha]

Hi All,

Can anybody please tell me how to propagate through the network to calculate the posterior probabilities in bayesian network when no evidence is provided or a single or multiple evidences are provided by the user?
Is there any step by step methodology to be followed to get the inferences?

Actually I had read abuot the junction tree propagation algorithm for exact propagation, but not getting exactly how to implement that, as I am new to this area. I have't found any illustration of the algorithm on internet, so I am finding it difficult to get along with it.

And also is it necessary to apply such algorithms to get the inferences for all nodes when the evidences are set?

[shooltz[BayesFusion]]

For notes on implementing algorithms based on junction tree, try "Inference in Belief Networks: A Procedural Guide" by Huang and Darwiche.

[megha]

For notes on implementing algorithms based on junction tree, try "Inference in Belief Networks: A Procedural Guide" by Huang and Darwiche.

Hi,
I had read out that, but tell me is it necessary to implement that algorithm for calculating inferences? And if we implement that then do we have to build the junction tree every time for calculating inferences each time we set the evidence or is it a offline process means we must develope the junction tree and save it for further use of inferencing?

[shooltz[BayesFusion]]

I had read out that, but tell me is it necessary to implement that algorithm for calculating inferences?

I'm not quite sure what's your actual question. JT-based algorithm is one (arguably most commonly used) way to implement exact BN inference, but there are other exact algorithms.

And if we implement that then do we have to build the junction tree every time for calculating inferences each time we set the evidence or is it a offline process means we must develope the junction tree and save it for further use of inferencing?

Both approaches will work.

[megha]

I had read out that, but tell me is it necessary to implement that algorithm for calculating inferences?

I'm not quite sure what's your actual question. JT-based algorithm is one (arguably most commonly used) way to implement exact BN inference, but there are other exact algorithms.

And if we implement that then do we have to build the junction tree every time for calculating inferences each time we set the evidence or is it a offline process means we must develope the junction tree and save it for further use of inferencing?

Both approaches will work.

Hi,

While developing junction tree algorithm we have to develope the moral graph for the bayesian network structure. While developing the moral graph for each node we have to consider the set of parents of that node and have to join every pair of parents, whatever may be the number of parents of that node, right?

[shooltz[BayesFusion]]

Sorry, this is SMILE support forum - your question is off-topic. We'll be glad to answer any questions related to SMILE/GeNIe.