摘要: 针对SEM算法易收敛于局部最优的问题,提出一种改进SEM算法——DSEM-PACOB算法,采用PACOB算法提供的良好候选网络及合理的策略,估计节点变量缺失值,并使与待估节点变量紧密相关的若干节点变量直接参与估计。实验结果表明,与SEM算法相比,利用新算法获得的最终解的质量有所提高,且具有更好的稳定性。
关键词:
学习贝叶斯网,
PACOB算法,
紧密相关,
新决策网络
Abstract: Aiming at the problem that SEM algorithm always converges to local optimal network, an improved SEM algorithm called DSEM-PACOB algorithm is proposed, which makes use of the better candidate networks and reasonable strategy provided by PACOB algorithm to estimate the missing value of node variables. Meanwhile, it chooses several node variables, which have close correlations with the estimated node variable, to conduct estimation. Experimental results show that, compared with SEM algorithm, this new algorithm makes qualitative improvements on the quality of the final solutions, and has better performance of stability.
Key words:
learning Bayesian network,
PACOB algorithm,
close correlation,
new deciding network
中图分类号:
廖学清;吕 强;单冬冬. 数据缺失下学习贝叶斯网的SEM算法[J]. 计算机工程, 2009, 35(8): 214-216.
LIAO Xue-qing; LV Qiang; SHAN Dong-dong. SEM Algorithm with Values Missed in Learning Bayesian Network[J]. Computer Engineering, 2009, 35(8): 214-216.