LBP convergence on simple graphi

[aj.sh...@gmail.com]

Hi,

I'm totally new to probabilistic graphical models and opengm.

I have implemented simple pairwise Markov network based on opengm samples and
PGM course instructed by professor Dophne Koller, the network structure attached.
The code hosted on this repository: https://github.com/alishir/opengm-quickstart

Energy calculation is fine, but when I using belief propagation to find most probable labling,
the inference algorithm doesn't converge. I also used TRBP algorithm but it cause run time
exception with following message.

OpenGM assertion mpParameter.specialParameter_[i] > 0 failed in file /usr/local/include/opengm/inference/messagepassing/messagepassing_trbp.hxx, line 120

Would you please help me to solve  problems in LBP convergence and TRBP exception?

Regards,
Ali