Software
Local bayesian network structure learning
using a constraint-based approach
Here below is a free software written in C by Sergio Rodrigues de Morais (Ph.D.)
that implements an efficient and scalable constraint-based algorithm
for local bayesian network structure learning, called Feature Subset Selection (FSS).
The software implements several procedures discussed in our articles (see
references) : 'HPC'
for extracting the parents and children of a target variable and
'MBOR' for learning the Markov boundary of a target.
FSS combine ideas from incremental and
divide-and-conquer methods in a principled and effective way,
while still being sound in the sample limit. Several synthetic and real-world data
sets of various sample sizes are provided. The most noteworthy feature of FSS is
its ability to handle large neighborhoods contrary to current
constraint-based algorithm proposals. Based on our experiments,
we observed that the number of calls to the statistical test,
en hence the run-time, is approximatively on the
order O(n^(1.09)), where n is the number of variables, on several synthetic data sets, and O(n^(1.21)) on a real
drug design characterized by 140,000 features. Therefore, FSS is very fast
and doesn't require much memory to handle thousands of variables.
FSS is distributed in the hope that it will be useful to researchers and bayesian netwok practitionners,
but without any warranty; without even the implied warranty of
merchantability or fitness for a particular purpose.
This program is free software; you can redistribute it and/or modify it under the terms of the
GNU General Public License as published by the Free Software Foundation.
Download FSS latest version
