带区间参数的贝叶斯网的近似推理

 用区间概率代替贝叶斯网中的点概率,将贝叶斯网扩展为带区间参数的贝叶斯网,使得贝叶斯网更具一般性.用规范概念的计算规则进行区间概率的计算,用Gibbs采样的近似推理法对带区间参数的贝叶斯网进行近似推理,计算后验概率.并给出近似推理的算法和实验,实验结果表明该方法是有效的.

The approximate reasoning of Bayesian Network appending interval probability

Bayesian Networks can been expanded to Bayesian Networks appending interval probability by using interval probability to replace the probability tables in Bayesian Networks.Canonical concept is good for interval probability computation,although it has some disadvantages comparing to intuitive concept.For getting posterior probability,the approximate reasoning(Gibbs sampling)was used,then the arithmetic analysis and experiment results of approximate reasoning were given,the method was proved to be feasible.