I tried to use Viterbi algorithm for DBN, but I couldn't.
Also, I couldn't find the detailed usage (or some
example code) in the manual.
Moreover, it appears that Viterbi with the Wrapper (high
API) doesn't work
(the output should be the most likely sequence over all
time slices but I can only get one MPE for the last
Another question is how to make two nodes share a
single CPD (in learning,
the statistics on these two nodes will be accumulated
to this CPD). In BNT,
this functionality is provided by equivalent classes.