Re: [UAI] question!!! [I'm looking for an algorithm ...]

From: Milan Studeny (studeny@utia.cas.cz)
Date: Thu Jun 20 2002 - 13:17:54 PDT

  • Next message: icdm02@kis.maebashi-it.ac.jp: "[UAI] Call for Tutorials (IEEE Data Mining 2002)"

     Dear Linda,
    I guess that perhpas Ross Shachter's Bayes-ball algorithm is what
    you need.

    I myself proposed a similar algorithm in more general context
    of chain graphs: see paragraph 5.2 of my paper
    Bayesian networks from the point of view of chain graphs,
    Proceedings of 14-th UAI conference (1998), pp. 496-503.
    You can get it from
     http://www.utia.cas.cz/user_data/studeny/c7.html
    Thus, you can use my algorithm adapted to the context of acyclic digraphs.

    I hope that somebody else can give you an exact reference to Shachter's
    paper if you need it.

     Regards from
                      Milan Studeny

    On Tue, 18 Jun 2002, Linda Smail wrote:

    > Hello,
    > I'm looking for an algorithm for the folowing problem :
    > given (I, G ) an acyclique graphe,
    > given J subset of I
    > I want to identify all vertex from J, which are :
    > 1) connect by direct connection
    > or
    > 2) connected by chains such that for all vertex i by which it passes, we have :
    > a) chains convergent towards i if the vertex i is in J
    > b) chains in series if the vertex i is not in J.
    > is anyone here aware this kind of algorithm?
    > Thanks for your help.
    > Lynda
    >
    >

    ------- End of Forwarded Message



    This archive was generated by hypermail 2b29 : Thu Jun 20 2002 - 13:18:24 PDT