BayesFusion Support Forum

Am I missing something? You can use DSL_ALG_ID_SHACHTER algorithm - it'll find the best policy at SMILE level. It's the equivalent of 'Find best policy' selection from the Network|Algorithm in GeNIe. The code in the tutorial invokes the default influence diagram algorithm DSL_ALG_ID_COOPERSOLVING (...

I've trained my network with a trainings set. Now I want to test it with the test set. Is there a possibility to import the complete test set and compute it or have I to choose the evidence for each node by hand? The closest thing would be using Data|Copy Cases command, which would create the set o...

RemoveOutcome will fail if you have just two states/choices - SMILE always keeps the network in consistent state, and discrete node with just one outcome certainly breaks the consistency. SetNumberOfOutcomes will fail if node has children. This is rather obscure method, mainly used by file I/O :) If...

ausi wrote:Is this possible?

It is not possible with current SMILE implementation.

I believe you were hit by a serious bug in the SMILE API design. DSL_intArray (and other DSL_xxxArrays) has NumItems and GetSize, which can report different values. In your case, DSL_intArray::GetSize is the method to call.

Don't ask why it was design this way - nobody knows

Christian wrote:Old Training Time: 1864 minutes
New Training time: 1348 minutes

If you're still interested in EM, most recent GeNIe build (2.0.3092.0) should perform even better.

Your data uses comma (not decimal dot) - GeNIe treats the values as text, not numbers, so discretization is not possible.

Now my question is how i should represent this data. Is the solution to see each value as another state, what means 60 states for each node, or should I divide them into intervals? The data looks continous, so you should discretize it. The choice of intervals depends of course on the problem domain.

parseval wrote:May be we can update this. Thank you.

Yes, you can update this. GeNIe/SMILE wiki is open to user edits.

On my machine VOI is calculated as -2.84e-014. This value is very close to zero and is probably a result of precision loss in VOI algorithm.

Can you post your model so we can investigate?

How many values are possible in one temporal node in SMILE/GeNIe The only limit is available memory, and this depends on the OS and hardware the program is running on. (or what is the max. network size to load it to GeNie)? Since GeNIe is a Win32 binary, it's user-mode address space is limited to 2...

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

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 imple...

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