《计算机应用研究》|Application Research of Computers

基于变权重迁移学习的BN参数学习算法

BN parameter learning algorithm based on dynamic weighted transfer learning

免费全文下载 (已被下载 次)  
获取PDF全文
作者 郭文强,徐成,肖秦琨,李梦然
机构 1.陕西科技大学 a.电子信息与人工智能学院;b.电气与控制工程学院,西安 710021;2.西安工业大学 电子信息工程学院,西安 710021
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2021)01-022-0110-05
DOI 10.19734/j.issn.1001-3695.2019.10.0600
摘要 针对小数据集条件下的贝叶斯网络(Bayesian network,BN)参数估计困难问题,提出了一种基于变权重迁移学习(DWTL)的BN参数学习算法。首先,利用MAP和MLE方法学习得到目标域初始参数和各源域参数;然后根据不同源域数据样本贡献的不同计算源权重因子;接着基于目标域样本统计量与小数据集样本阈值的关系设计了目标域初始参数和源域参数的平衡系数;最后,基于上述参数、源权重因子和平衡系数计算得到新的目标参数。在实验研究中,通过对经典BN模型的参数学习问题验证了DWTL算法的有效性;针对小数据集下的轴承故障诊断问题,相较于传统迁移学习(LP)算法,DWTL算法学习精度提高了10%。实验结果表明:所提出的算法能够较好地解决样本数据集在相对稀缺条件下的目标参数建模问题。
关键词 小数据集; 贝叶斯网络; 迁移学习; 参数学习
基金项目 国家自然科学基金资助项目(61271362,62071366)
陕西省科技厅重点研发计划资助项目(2020SF-286)
陕西省科技厅自然科学基金资助项目(2017JM6057)
陕西省教育厅产业化研究项目(18JC003)
西安市科技计划资助项目(2019216514GXRC001CG002GXYD1.1)
本文URL http://www.arocmag.com/article/01-2021-01-022.html
英文标题 BN parameter learning algorithm based on dynamic weighted transfer learning
作者英文名 Guo Wenqiang, Xu Cheng, Xiao Qinkun, Li Mengran
机构英文名 1.a.School of Electronic Information & Artificial Intelligence,b.School of Electrical & Control Engineering,Shaanxi University of Science & Technology,Xi'an 710021,China;2.School of Electronic Information Engineering,Xi'an University of Technology,Xi'an 710021,China
英文摘要 To solve the problem of Bayesian network(BN) parameter estimation accuracy under small dataset conditions, this paper proposed a parameter dynamic weighted transfer learning algorithm(DWTL) based on varying weight transfer learning. Firstly, the presented algorithm used MAP and the MLE method to learn the initial parameters of the target domain and the parameters of each source domain. Then, it obtained the source weight factors of the source domain by the different data source contributions. Based on the sample statistic, this method helped to obtain the final target parameters by fusing the data size threshold values, the balance coefficients for the target initial parameters with the source domain parameters. The experimental results show that under the condition of the small data set, the learning accuracy of DWTL algorithm is better than MLE algorithm, MAP algorithm and traditional transfer learning algorithm(LP). Under the condition of sufficient data set, the learning accuracy of DWTL algorithm approaches the classical MLE algorithm, and verifies the correctness of the algorithm. Moreover, it demonstrates successful application to real-world bearing fault diagnosis case studies. Comparing with the LP algorithm, the DWTL algorithm achieves about 10% enhancement for the average diagnosis precision.
英文关键词 small dataset; Bayesian network; transfer learning; parameter learning
参考文献 查看稿件参考文献
 
收稿日期 2019/10/24
修回日期 2019/12/16
页码 110-114
中图分类号 TP311
文献标志码 A