A Probabilistic Relaxation Framework for Learning Bayesian Network Structures from Data

Graphical models have been very promising tools that can effectively model uncertainty, causal relationships, and conditional distributions among random variables. This work proposes a new probabilistic method for learning Bayesian network structures from data. In the proposed method the existence of an edge in the network is no longer considered as a hard or deterministic issue, but rather we assign a certain probability for the existence of each edge. The proposed method uses a global optimization approach, originally developed for clustering and classification problems, to find the set of edges probability that lead to the best network structure. The experimental results show that the proposed approach achieves very promising results compared to other structure learning approaches.