Closure of independencies under Graphoid properties: some experimental results

In this paper we describe an algorithm for computing the closure with respect to graphoid properties of a set of independencies. Since the computation of the complete closure is infeasible, we provide a procedure, called FC1, which is based on a unique inference rule and on the elimination of redundant independencies. FC1 is able to compute a reduced form of the closure, called fast closure, which is equivalent to the complete closure, but whose size is much smaller. Some experimental tests have been performed with an implementation of the procedure in order to show the computational behavior of the algorithm. We have also compared the computational cost and the size of the fast closure with the corresponding data for the complete closure.