信息不完备小样本条件下离散DBN参数学习  

Discrete dynamic BN parameter learning under small sample and incomplete information

作　　者：任佳[1] 高晓光[2] 白勇[1] 

机构地区：[1]海南大学信息科学技术学院,海南海口570228 [2]西北工业大学电子信息学院,陕西西安710129

出　　处：《系统工程与电子技术》2012年第8期1723-1728,共6页Systems Engineering and Electronics

基　　金：国家自然科学基金(61062006);天津大学-海南大学创新基金资助课题

摘　　要：针对信息不完备小样本条件下离散动态贝叶斯网络参数学习问题,提出约束递归学习算法。该方法通过前向算法建立含有隐藏变量的离散动态贝叶斯网络参数递归估计模型,以当前时刻网络参数为变量,构建均匀分布表示的先验参数约束模型。在此基础上利用优化算法获得近似的Beta分布,将该分布下的先验参数信息加入递归估计模型中完成参数学习。通过无人机动态威胁评估模型验证了该方法的有效性和精确性。Aiming at the discrete dynamic Bayesian network parameter learning under the situation of small sample and incomplete information, a constraint recursion learning algorithm is presented. The forward algo- rithm is used to establish a parameter recursion estimation model of discrete dynamic Bayesian network with hid- den variables. A prior parameter constraint model with uniform distribution is established with the present net- work parameters as variables. Then the approximate Beta distribution could he acquired through the optimiza- tion algorithm. Finally, the distribution of prior parameter knowledge could be used in the above model of recur- sire estimation to finish the parameter learning process. The method is applied to the unmanned aerial vehicle dynamic model of threat assessment. The results show the effectiveness and accuracy of the proposed algo- rithm.

关 键 词：离散动态贝叶斯网络 参数学习 约束递归学习 信息不完备 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]