基于根节点优先搜索的信度传输DBP算法研究

Research on Traffic Situation Assessment Modeling Based on DBP Algorithms

针对信度传输算法迭代次数较多的问题,提出一种基于根节点优先搜索的信度传输DBP算法。DBP算法依据根节点优先搜索的原理,选择一种特定的节点顺序进行信度传播,直接到达信息传播的不动点,降低迭代次数,节省推理时间。首先,分析了BP算法的主要思想、工作原理及推理过程,其次,提出了DBP算法,建立了贝叶斯网络模型,给出了该算法的基本原理,最后,给出了DBP算法流程,并通过典型的树形结构的贝叶斯网络实例,对DBP算法进行了分析,结果表明DBP算法在推理时间上优于BP算法,算法的时间优化率更高,从而验证了DBP算法的有效性。

To the question of multi-iterative degree,a new belief propagation algorithm Based on First Search of root node is proposed,According to the principle of deepness first search of tree’s root node,DBP algorithm selects a specially appointed subsequence to be⁃lief propagation,reaches directly immobility node of belief propagation,thus,reduces the iterative degree and inference time.Firstly,the main idea,work principle and inference process are summarized;secondly,DBP algorithm is proposed,Bayesian Network model is es⁃tablished,the basic principle is given;finally,the algorithm flow are given,by some simulation of typical Bayesian Network of tree,DBP algorithm is analyzed detail.Experimental results show that DBP algorithm is more effective and efficient in inference time,thus the va⁃lidity of DBP algorithm is verified.