Difference between revisions of "Relational Markov network"
From Cohen Courses
Jump to navigationJump to search (Created page with 'A relational Markov network (RMN) is a model for data with relations and discrete attributes. It is specified by a set of clique templates <math>C</math> and corresponding potent…') |
|||
| Line 1: | Line 1: | ||
| − | A relational Markov network (RMN) is a model for data with relations and discrete attributes. It is specified by a set of clique templates <math>C</math> and corresponding potentials <math>\Phi</math>. Given a relational database, the RMN produces an unrolled Markov network over all the attributes <math>X</math> . The cliques <math>c</math> instantiated by a template <math>C</math> share the same clique potential . The combined probabilistic model is <math>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. | + | A relational Markov network (RMN) is a model for data with relations and discrete attributes. It is specified by a set of clique templates <math>C</math> and corresponding potentials <math>\Phi</math>. Given a relational database, the RMN produces an unrolled Markov network over all the attributes <math>X</math> . The cliques <math>c</math> instantiated by a template <math>C</math> share the same clique potential . The combined probabilistic model is <math>p(X) = \frac{1}{Z}\Pi_{c\in C}\Phi_c(X(c))</math>. In general, a template can be any boolean formula over the relations. |
Latest revision as of 06:14, 4 October 2012
A relational Markov network (RMN) is a model for data with relations and discrete attributes. It is specified by a set of clique templates and corresponding potentials . Given a relational database, the RMN produces an unrolled Markov network over all the attributes . The cliques instantiated by a template share the same clique potential . The combined probabilistic model is . In general, a template can be any boolean formula over the relations.
