Changing How Messages are Calculated in Belief Propagation

[Adam]

Hello,

Would it be possible to use this BP code where I define my own function to calculate the messages? For instance, I would like to run min-sum BP where each factor -> variable message is calculated by running an algorithm like simulated annealing as opposed to a brute force search or whatever the current implementation may be.

Thank You!