I’m curious if there is a trick to make this problem convex.
I have a convex objective when the domain is a probability distribution on random variables X,Y,Z (all alphabets are finite).
I would like to represent a Markov relation, say X-Y-Z, between the random variables as a linear constraint so it fits the convex optimization model.
In case you are not familiar with Markov chains, the Markov X-Y-Z implies that p(x|y,z) = p(x|z) for all x,y,z.
Thanks in advance,