DBN条件下信度传输优化算法仿真研究

Simulation Research on Belief Propagation Optimization Algorithm under DBN Conditions

信度传输算法是一种对贝叶斯网络等图模型进行推理的消息传递算法。针对信度传输算法迭代次数较多的问题,提出一种信度传输优化算法。首先,分析了BP算法的推理过程;其次,提出了优化的信度传输算法,分析了树的定义及树的遍历,给出了优化的信度传输算法的基本原理和实现步骤;最后,在动态贝叶斯网络DBN条件下,通过单一证据和组合证据推理,对优化的DBP算法与BP算法进行实验研究。结果表明:DBP算法可减少迭代次数,节省更多的推理时间,提高算法有效性。

Belief propagation algorithm is a message transfer algorithm which infers the contour model of Bayesian network.To the problem of multi-iterative degree,belief propagation optimization algorithm is proposed.Firstly,the reasoning process of BP algorithm is analyzed.Secondly,the optimized belief propagation algorithm is proposed.The definition of tree and the traversal of tree are analyzed.The basic principle and implementation steps of the optimized belief propagation algorithm are given.Finally,under the condition of dynamic Bayesian network(DBN),through single evidence and combined evidence reasoning,the optimized DBP algorithm and BP algorithm are studied experimentally.The results show that DBP algorithm can reduce the number of iterations,save more reasoning time,and improve the effectiveness of the algorithm.