基于时间窗的SVDBN近似推理算法研究

变结构动态贝叶斯网络(SVDBN)在处理非稳态过程的不确定问题具有其独特的优越性。为克服SVDBN推理算法不能实现在线推理的缺陷,在引入SVDBN的时间窗和时间窗宽度概念基础上,阐述了基于时间窗的SVDBN在线近似推理机制,提出了2种在线近似推理算法,即基于时间窗的变结构离散动态贝叶斯网络(SVDDBN)递推推理算法和基于时间窗的SVDDBN快速推理算法。通过仿真实验验证了这2种算法的有效性,并从复杂度、适用范围及更新时间等方面进行了比较。

Research on Approximate Reference Algorithm of SVDBN based on Sliding Window

Structure-variable dynamic Bayesian networks (SVDBN) have the special advantage in dealing with the uncertainty of the unstable processes. In order to overcome the disadvantage that the inference algorithms of the SVDBN are unable to apply online, introducing the concepts of SVDBN sliding window and the window width, the online approximate inference mechanism of structure-variable dynamic Bayesian networks based on sliding window is explained, and two online algorithms are proposed, that is the recursive inference algorithm of structure-variable discrete dynamic Bayesian networks (SVDDBN) based on sliding window and the fast inference algorithm of SVDDBN based on sliding window. Experimental simulations show the effectiveness of the two inference algorithms and compare their complexity, application, updated time and so on.