Controlling the Inconsistent of the Bayesian Network Structure Learning with the Recursive Autonomy Identification
Renqing Duan, Youlong Yang, Guozhou Li · 2014
In the constraint-based Bayesian Network structure learning algorithms, many of them suffer from statistic errors in conditional independence tests. Due to the recursive autonomy identification algorithm combining the conditional independence tests and edges direction from the outset and along the procedure, appearing the inconsistence v-structures is frequent. In this paper, we propose an algorithm which embeds an controlling the inconsistence v-structures procedure in the orientation stage of recursive autonomy identification algorithm. It is efficient to avoid the inconsistence v-structure. We show the advantages of the proposed algorithm by comparing with RAI, PC, SCA and MMHC over the structure correctness and algorithm complexity.