Research Article

Upper-Lower Bounds Candidate Sets Searching Algorithm for Bayesian Network Structure Learning

Algorithm 1

UGLA.
(1) Input: Data set D; Variable set ;
(2) Initialization: undirected graph , where ;
(3) Order-0 CI test: for each pair variables , if is hold, then , where
   is an undirected edge;
(4) For each triplet variables
(5)     For each pair variables , and , if , , , and is hold,
   then ,
(6) End for;
(7) Triangulate undirected graph , what results with
(8) Call the junction tree algorithm to obtain maximal prime sub-graphs from ;
(9) Output and as final result.