A novel scalable and correct Markov boundary learning algorithm under faithfulness condition

Abstract

In this paper, we propose a novel constraint-based Markov boundary discovery algorithm, called MBOR, that scales up to hundreds of thousands of variables. Its correctness under faithfulness condition is guaranteed. A thorough empiric evaluation of MBOR’s robustness, efficiency and scalability is provided on synthetic databases involving thousands of variables. Our experimental results show a clear benefit in several situations: large Markov boundaries, weak associations and approximate functional dependencies among the variables.

2 Figures and Tables

Cite this paper

@inproceedings{Morais2008ANS, title={A novel scalable and correct Markov boundary learning algorithm under faithfulness condition}, author={Sergio Rodrigues de Morais and Alex Aussem}, year={2008} }