Hello,

I am trying to segment an MTS whilst learning DBNs within it.
I am using a hidden variable to model the changes in segment and the SEM
algorithm to learn the parameters and structure.

However, I find that the SEM does not converge to find the correct
segments.

I am wondering if anyone has tried anything similar to this before and
encountered any problems. I am using inference to calculate the expected
statistics.

Thanks,

Allan Tucker,
Birkbeck College,
University of London.


Reply via email to