Information Technology Reference
In-Depth Information
EvoBANE uses a general-purpose grammar-guided programming core that im-
plements a complete evolutionary mechanism (initialization, selection, crossover
and replacement) whose grammatical crossover operator avoids the closure
problem.
A CFG generator has been implemented in EvoBANE in order to automate
the creation of context-free grammars that generate languages whose sentences
codify valid Bayesian network structures that preserve DAG constraints. The
CFG generator inputs the specifications of the classification problem to solve
and the features of the solutions to be generated, and outputs the CFG that
EvoBANE uses to generate and evolve a population of Bayesian networks.
The results show that the Bayesian networks automatically generated by
EvoBANE accurately solve classification problems from two different applica-
tion domains. In these two cases EvoBANE has been able to explore the search
space, avoiding local optima and reaching good Bayesian networks without a
repair mechanism. The flexibility of the CFG generator vests EvoBANE with
the capability to modify the search space in order to find the simplest Bayesian
network that solves the problem.
Acknowledgements
This work has been supported by the LIA-Group (http://www.lia.upm.es/) of
the Universidad Politécnica de Madrid.
References
1. Wong, M.L., Guo, Y.Y.: Learning bayesian networks from incomplete databases
using a novel evolutionary algorithm. Decision Support Systems 45, 368-383 (2008)
2. Darwiche, A.: Bayesian networks. Communications of the ACM 53(12), 80-90
(2010)
3. Niedermayer, D.: An Introduction to Bayesian Networks and Their Contemporary
Applications. In: Innovations in Bayesian Networks. SCI, pp. 117-130. Springer,
Heidelberg (2008)
4. Chen, X., Anantha, G., Lin, X.: Improving bayesian network structure learning
with mutual information-based node ordering in the k2 algorithm. IEEE Transac-
tions on Knowledge and Data Engineering 20(5), 1-13 (2008)
5. Larrañaga, P., Poza, M., Yurramendi, Y., Murga, R.H., Kuijpers, C.M.H.: Struc-
ture learning of bayesian networks by genetic algorithms: A performance analysis
of control parameters. IEEE Transactions on Pattern Analysis and Machine Intel-
ligence 18(9), 912-926 (1996)
6. Barriére, O., Lutton, E., Wuillemin, P.: Bayesian network structure learning us-
ing cooperative coevolution. In: GECCO 2009, Montréal, Canada, July 2009,
pp. 755-762 (2009)
7. Alonso-Barba, J.I., delaOssa, L., Puerta, J.M.: Structural learning of bayesian net-
works by using variable neighbourhood search based on the space of orderings. In:
Ninth Conference on Intelligent Systems Design and Applications, pp. 1435-1440
(2009)
Search WWH ::
Custom Search