一类贝叶斯网络的线性推理

The Linear Reasoning of Bayesian Networks

贝叶斯网络提供了表示变量集之间概率依赖性的一个自然有效的方法,而且其推理方法是主观贝叶斯方法的一个扩展,具有坚实的概率理论基础,因此,许多人工智能的研究者都采用贝叶斯网络作为一种知识表示的方法,将其应用到各种问题领域。如:故事理解、规划、电路错误检测和医学诊断等等。但是,贝叶斯网络已遭受到一些人工智能研究者的批评,因为它们需要大量的数值概率值使不确定关系量化,

Uncertain reasoning is one of the key technologies in intelligent systems and diagnosis systems. Probably one of the most popular models for uncertain reasoning is Bayesian networks, but the computational complexity probabilistic inference using belief networks is NP-hard. In view of it, two theorems are given in this paper. We substitute a linear function for the conditional probability matrix to propagate probabilities in Bayesian networks so that the reasoning speed is improved.