Re: [UAI] mixed variables

From: Ellis Clarke (clarke@cs.umbc.edu)
Date: Mon Apr 10 2000 - 09:12:14 PDT

  • Next message: Francisco Levinson: "[UAI] 2001 WSES International Conference on:"

      Since most algorithms for Bayesian Belief Network (BBN) creation require
    discrete data, a method to convert continuous data into discrete is used
    in a pre-processing step. An information-theoretic metric is used in a
    recently published article by myself and Bruce Barton. The discretization
    method can be extended to dynamically combine existing partitions when
    creating a BBN, if a Minimum Descriptive Length (MDL) metric is used to
    guide the BBN creation.
      This method does not rely on assumptions about the data distribution and
    is intentionally designed to handle 'multi-modal' data distributions with
    a minimal loss of information.
     The reference is:

      Clarke,E., and Barton,B., (2000), Entropy and MDL Discretization of
    Continuous Variables for Bayesian Belief Networks. International Journal
    of Intelligent Systems, 15, 61-92.

      Another relatively recent article on the same topic is:

      Monti,S., and Cooper,G., (1998), A Multivariate Method for Learning
    Bayesian Networks from Mixed Data, Proc. Uncertainty in Artificial
    Intelligence, ed. Cooper,G., and Moral,S., Morgan Kaufmann, S.F.

      I hope this helps.

      Ellis

    On Thu, 6 Apr 2000, Zhu wrote:

    > Hello all,
    >
    > Is there any algorithms of Bayesian Network to work directly on the
    > mixture of continuous and categorical variables?
    >
    > The classification problem that I am working on has 37 input variables, 15
    > of them are categorical and the rest of them are continuous. To my
    > understanding, I need to discretize the continuous varibles in order to
    > apply some commonly used algorithms (such as junction tree) to construct
    > and estimate BNs. Since a large portion of the input variables are
    > continuous, I am afraid of loss of information by discretizing them.
    > References and input on working directly on the mixture will
    > be highly appreciated. I would also like to have any comments and
    > experiences on how much gain we can get from working on the mixture
    > directly over transforming all variables into discrete. Thanks.
    >
    >
    > Best regards,
    >
    > Julie
    >
    >

    _____________________________________________________________________
      Ellis Clarke, Ph.D.; CSEE, University of Maryland Baltimore County;
      clarke@cs.umbc.edu
    _____________________________________________________________________



    This archive was generated by hypermail 2b29 : Mon Apr 10 2000 - 09:17:43 PDT