An algorithm that does just that is described in a paper by Christopher Meek,
"Causal Inference and Causal Explanation with Background Knowledge," which
appears in the UAI-95 proceedings, on pp.403-410. The input of this algorithms
includes, among other things, a specification of forbidden edges (set F in the
paper) and required edges (set R in the paper). Unfortunately, there is an
error in that algorithm that causes it to produce extensions (DAGs) that do not
conform to the background knowledge. (When shown the error, Meek recognized
it in e-mail correspondence with me.) In particular, it is possible for the
final DAG to contain an edge that was listed as forbidden. The error was
found by my student, Bhaskara Moole. Error and correction are described in
his M.S. thesis ("Parallel Construction of Bayesian Belief Networks").
The correction consists simply in the addition of Step S1 of Phase II'' as
the first step (before S1) of Phase III in Meek's algorithm.
>
> Thanks in advance,
> Uschi Robers
>
>
You are welcome.
Marco Valtorta, Associate Professor and Undergraduate Director
Department of Computer Science internet: mgv@usceast.cs.sc.edu
University of South Carolina tel.: (1)(803)777-4641 fax: -3767
Columbia, SC 29208, U.S.A. http://www.cs.sc.edu/~mgv/ tlx: 805038 USC