Hi everybody

I have to write my own factor. I have read the documentation for creating a new factor and operators. I don't understand how I can define messages which should be passed through edges. I studied EP in detail but I can't find the right messages to be passed
in my own factor. As far as I understood, in EP message passing, the value for each variable node should be approximated using the posterior computed in the factor node in each iteration. That's like computing the inverse of factor function and re-distribute
the value again to the variable nodes. Am I right?

Would you please give me some clue in this case? How can I define the messages? My Operation is a complicated one, with four parameters. These parameters have not an equal role in computing the factor. For example one is a power and one is a coefficient.What
should I do in this case for defining messages?

Any help is appreciated!