计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (17): 128-133.

• 数据库、数据挖掘、机器学习 • 上一篇    下一篇

单元化单隐变量变结构DDBN推理算法

陈海洋,毛蕊蕊,聂弘颖   

  1. 西安工程大学 电子信息学院,西安 710048
  • 出版日期:2015-09-01 发布日期:2015-09-14

Unitized single hidden variables structure-variable DDBN inference algorithm

CHEN Haiyang, MAO Ruirui, NIE Hongying   

  1. School of Electronic Information, Xi’an Polytechnic University, Xi’an 710048, China
  • Online:2015-09-01 Published:2015-09-14

摘要: 变结构动态贝叶斯网络(DBN)描述的是一个非稳态随机过程,是一种更灵活、更有效的动态网络。为了克服现有变结构DBN的推理算法不能实现在线推理的缺陷,提出了一种近似在线推理算法——单元化单隐变量变结构离散DBN(DDBN)推理算法。在定义了单隐变量变结构离散动态贝叶斯模型和单元的基础上,提出了算法的基本思想,并从理论上对算法进行了推导。仿真实验验证了该算法的正确性和有效性。

关键词: 信息传播, 变结构离散动态贝叶斯网络, 近似推理算法, 不确定性

Abstract: Structure-variable Dynamic Bayesian Network (DBN) is used to describe an unstable process and is a kind of more flexible and effective network. In order to overcome the disadvantage that the inference algorithm on structure-variable dynamic Bayesian networks can’t infer online, an approximate online inference algorithm is proposed, that is, the unitized single hidden variable structure-variable Discrete DBN (DDBN) inference algorithm. On the base of defining the single hidden variable structure-variable DDBN model and the unit structure, it proposes the basic idea of the algorithm, and deduces the algorithm in theory. It is proved by the simulation experiments that this algorithm is correct and efficient.

Key words: information dissemination, structure-variable Discrete Dynamic Bayesian Network (DDBN), approximate inference algorithm, uncertainty