Relational Markov network

A relational Markov network (RMN) is a model for data with relations and discrete attributes. It is specified by a set of clique templates $C$ and corresponding potentials $\Phi$. Given a relational database, the RMN produces an unrolled Markov network over all the attributes $X$ . The cliques $c$ instantiated by a template $C$ share the same clique potential . The combined probabilistic model is $p(X)={\frac {1}{Z}}\Pi _{c\in C}\Phi _{c}(X(c))$. In general, a template can be any boolean formula over the relations.