离散动态贝叶斯网络的直接计算推理算法

Direct calculation inference algorithm for discrete dynamic bayesian network

离散动态贝叶斯网络是对动态过程进行建模和定性推理的有力工具。但是目前所用的各种推理算法都需要进行复杂的图形变换,不易于计算机编程实现而且计算时间长。为此,基于概率论和贝叶斯网络的基本性质,提出了离散动态贝叶斯网络的直接计算推理算法,从理论上对算法进行了推导并进行了实例验证。该算法的最大优点就是不需要复杂的图形变换,非常适合于计算机编程实现,而且在某些情况下推理速度快于其它算法。

Discrete dynamic Bayesian network is a capable tool for modeling and quality inferring for dynamic process, but the current inference algorithms we can read in all materials are all based on complicated figure transformations. They are hard to programming and need long time for calculation. Aimed on this problem, we present a direct calculation inference algorithm for discrete dynamic Bayesian network based on the probability theory and the basic characters of the Bayesian network and verify it by samples. The most advantage of this algorithm is that it needs not performing complicated figure transformation, it is easy to programming, and under some conditions, it can work out the results quickly and directly. It is more useful in some applications where the time requests is relaxed.