Emergence of optimal Bayesian networks from datasets without backtracking using an evolutionary algorithm

Authors:
Isaac O. Osunmakinde;Anet Potgieter
Affiliations:
University of Cape Town, Cape Town, Republic of South Africa;University of Cape Town, Cape Town, Republic of South Africa
Venue:
CI '07 Proceedings of the Third IASTED International Conference on Computational Intelligence
Year:
2007

Citing 9
Cited 1

Probabilistic reasoning in intelligent systems: networks of plausible inference

Probabilistic reasoning in intelligent systems: networks of plausible inference
Learning belief networks from data: an information theory based approach

CIKM '97 Proceedings of the sixth international conference on Information and knowledge management
Artificial intelligence: a new synthesis

Artificial intelligence: a new synthesis
A Permutation Genetic Algorithm For Variable Ordering In Learning Bayesian Networks From Data

GECCO '02 Proceedings of the Genetic and Evolutionary Computation Conference
Artificial Intelligence: A Modern Approach

Artificial Intelligence: A Modern Approach
Information Theory, Inference & Learning Algorithms

Information Theory, Inference & Learning Algorithms
Exact Bayesian Structure Discovery in Bayesian Networks

The Journal of Machine Learning Research
Large-Sample Learning of Bayesian Networks is NP-Hard

The Journal of Machine Learning Research
Learning Bayesian network structures by searching for the best ordering with genetic algorithms

IEEE Transactions on Systems, Man, and Cybernetics, Part A: Systems and Humans

Emergent future situation awareness: a temporal probabilistic reasoning in the absence of domain experts

ICANNGA'09 Proceedings of the 9th international conference on Adaptive and natural computing algorithms

Quantified Score

Hi-index	0.01

Visualization

Abstract

We propose a new Hybrid Genetic Algorithm (HGA) developed from the domain of evolutionary algorithms to evolve optimal Bayesian networks from datasets. For its learning process, it uses genetic operators engineered from information theoretic and mathematical fields including Mutual Information (MI), Extended Dependency Analysis (EDA), Mathematical Power Sets and Minimum Description Length (MDL). Unlike our HGA, existing genetic algorithms (GAs) use genetic operators that usually use backtracking, which is an overhead for a learning algorithm. In our research, we prevented backtracking using an inner-loop and carried out several evaluation experiments. Our empirical results and structural evaluations showed that a HGA can discover optimal networks from datasets that we selected from different domains.