Energy minimization with constraints on configurations

[Erik Conser]

Hello,

I am using explicit functions and belief propagation to find the minimum energy configuration of a factor graph.  I've encountered a situation where I'd prefer to have the minimum energy configuration subject to certain constraints, rather than the absolute minimum energy.  Specifically, when two variables are using the same explicit function, I want the minimum energy configuration where they are using different labels; configurations where they use the same label should be ignored.  I don't see anything in the documentation that would allow me to do this, would I need to write my own inference code?  Any advice on this would be appreciated.

Thanks,

Erik Conser

[Oliver Mader]

Hej Erik,

one way to do that is to introduce a binary factor for both your variables and specify inf energies for the configurations you want to disallow and a constant for all the others. Note however, that this might change the choice of inference strategy.

Bests

Oliver