Problem with time count.

[nirzhor26]

Hi,
Does GeNIe have any problem in counting higher time slices? I want to analyse DBN model of 9 nodes (2 state in each node) for 500 time slices. But the window is showing the network is too complex after 42 time count.

[shooltz[BayesFusion]]

But the window is showing the network is too complex after 42 time count.

-42 is the error code associated with the 'network is too complex message'. GeNIe's inference in dynamic networks is based on the concept of the network unrolling: the program creates the temporary static network based on the dynamic network. The slice count parameter controls the size of the unrolled network. Setting it to 500 may result in a model which is too complex for inference algorithm to succeed.

If you'e only interested in posteriors of some subset of nodes, try marking them as targets by right-clicking and selecting 'Set Target' option from the context menu. This may help in reducing the complexity.

Also, what is the highest time point you're setting evidence at? 500 slices is rather large number.

[marek [BayesFusion]]

I just wanted to add that you may want to see the posteriors over a smaller number of slices. In my experience, the posteriors over different dime slices very quickly converge and don't change much. Perhaps 500 slices is a little too much? 50 slices may show you what you want. Please try "Temporal Beliefs" in the "View" menu of a DBN.
Cheers,

Marek