In this paper, we firstly formulate the concept of “ordering of sets” to represent the relationships between classes of variables. And then a parallel algorithm with little inter-processors communication is proposed based on “ordering of sets”. In our algorithm, the search space is partitioned in an effective way and be distributed to multi-processors to be searched in parallel. The results of experiments show that, compared with traditional greedy DAG search algorithm, our algorithm is more effective, especially for large domains.
Weitere Kapitel dieses Buchs durch Wischen aufrufen
- Parallel Learning of Bayesian Networks Based on Ordering of Sets
S. S. Zhang
- Springer Berlin Heidelberg