论文部分内容阅读
贝叶斯网络是不确定性环境下知识表示和推理的有效工具之一。现有的贝叶斯网络结构学习算法不同程度地存在学习效率偏低的问题,为此,本文提出一种高效而且可靠的贝叶斯网络结构学习算法ISOR。首先使用最大生成树算法和启发式切割集搜索算法以确定网络中所有可能的边,然后结合碰撞识别方法和启发式打分-搜索方法识别出所有边的方向,最后进行冗余边检验。与当前基于依赖分析的其它算法相比,该算法有效降低条件独立性检验的次数和阶数。算法分析和应用于Alarm网络的实验结果均表明,算法ISOR具有良好的性能。
Bayesian network is one of the most effective tools for knowledge representation and reasoning in uncertain environment. The existing Bayesian network structure learning algorithm has the problem of low learning efficiency to varying degrees. To this end, this paper presents an efficient and reliable Bayesian network structure learning algorithm ISOR. First, we use the maximum spanning tree algorithm and the heuristic cutting-set search algorithm to determine all possible edges in the network, and then identify the directions of all the edges by using the collision recognition method and the heuristic scoring-search method, and finally perform the redundancy edge test. Compared with other current algorithms based on dependency analysis, this algorithm effectively reduces the number and order of conditional independence tests. The analysis of the algorithm and the experimental results applied to the Alarm network show that the algorithm ISOR has good performance.