How can I close a half edge with a constant factor?

[meditans]

When learning about factor graphs, it's often said that one can close a half-edge, say on the variable x, by inserting a constant factor node f(x) = 1 (of course, this doesn't change the overall distribution). But I don't know how to do that in RxInfer. For example, I would write:

x ~ Bernoulli(p)

and, supposing x is a half-edge that I want to close, I will often write:

x ~ Bernoulli(0.5)

But this seems semantically different from just adding the constant factor in the graph. Is it? How should I do this correctly?

[bvdmitri]

You can do that with

x ~ Uninformative()

which will send a missing object as a message. The idea here is that missing represents f(x) = 1 Our product rules defined such that missing * any_other_message equals any_other_message. But, if this missing message goes to another node you will encounter "rule not found error". Works well with mean-field constraints though

[bvdmitri]

Oh wait, I actually was wrong, is an input to any node (either message or a marginal) is missing than the resulting message is also missing