The BNT Soft Discretization package is an extension
for Matlab's Bayes Net Toolbox BNT: This Soft Discretization package provides the following
functionality:
Implements soft discretization to train the CPTs of a discrete
Bayesian
Network directly from continuous data.
Allows you to use soft discretization also for inference and to
convert the inference results from the discrete network to meaningful
continuous output values.
The training method consists of a soft discretization step that
converts the
continuous variables of the training cases into soft evidence, followed
by a suitable parameter learning algorithm for the Bayesian
Network.
Most existing soft discretization approaches for Bayesian Networks use
fuzzy set theory which is based on membership functions. In
contrast out method starts out with a probability density function that
spreads the influence of a continuous variable to its neighbors,
followed by a discretization step. Thus our approach to soft
discretization is based on probability theory, rather than fuzzy set
theory. It turns out that a membership function can be
generated from the probability density function through convolution,
yielding a set of probability-based membership functions.
Documentation:
GT-ME-2009-002.pdf
(Technical Report -- Imme Ebert-Uphoff,
"A Probability-Based Approach to Soft
Discretization for Bayesian Networks", Research Report
GT-ME-2009-002, Georgia Institute of Technology, School of Mechanical
Engineering, Sept 22,
2009.) Derives the algorithms
for soft discretization.