sgfrazie
I'm curious.  In the contingency table fit, how does BayesiaLab calculate the entropy of a fully connected network.  I'm assuming it does not compose a fully connected network in the "learning" process for the calculation.
Quote 0 0
Dan
Indeed, the fully connected network is not necessary for this computation, you just have to index the data (i.e. identify all the different combinations of values) and then compute the frequency of each index to get the entropy of the data.

Hope this clarifies
Quote 0 0
sgfrazie
Thank you Dan.  Makes sense - simply apply -p*log(p) to each of the indexes to determine entropy of fully connected network. 
Quote 0 0