| E‐MS算法的收敛性 |
| |
| 引用本文： | 徐平峰,陈婷,尚来旭.E‐MS算法的收敛性[J].吉林大学学报(理学版),2019,57(5):1127-1130. |
| |
| 作者姓名： | 徐平峰 陈婷 尚来旭 |
| |
| 作者单位： | 长春工业大学数学与统计学院,长春,130012;长春工业大学数学与统计学院,长春,130012;长春工业大学数学与统计学院,长春,130012 |
| |
| 基金项目： | 国家自然科学基金;国家自然科学基金;国家自然科学基金;吉林省教育厅科研项目 |
| |
| 摘 要： | 考虑E MS算法的收敛性. 首先, 给出观测广义信息准则(GIC)最小值点的必要条件； 其次, 在模型空间有限性、 参数空间紧性、 Q函数连续性的条件下, 证明E MS算法产生序列的极限点满足观测GIC最小值点的必要性, 是对E MS算法全局收敛性的补充； 再次, 给出满足该必要条件但不满足全局收敛条件高斯图模型的一个实例； 最后, 证明E MS算法的全局收敛性.
|
| 关 键 词： | 缺失数据 模型选择 观测GIC E-M S算法 |
| 收稿时间： | 2018-10-22 |
Convergence of E-MS Algorithm |
| |
| XU Pingfeng,CHEN Ting,SHANG Laixu.Convergence of E-MS Algorithm[J].Journal of Jilin University: Sci Ed,2019,57(5):1127-1130. |
| |
| Authors: | XU Pingfeng CHEN Ting SHANG Laixu |
| |
| Institution: | School of Mathematics and Statistics, Changchun University of Technology, Changchun 130012, China
|
| |
| Abstract: | We considered the convergence of E MS algorithm. Firstly, we gave a necessary condition for the minimum point of observed generalized information criterion (GIC). Secondly, under the conditions of finiteness of model space, compactness of parameter space, continuity of Q function, we proved that it was necessary for the limit points of the sequence generated by E MS algorithm to satisfy the minimum points of observed GIC, which was a supplement to the global convergence of E MS algorithm. Thirdly, we gave an example of Gaussian graphical model which satisfied the necessary condition, but did not satisfy conditions for global convergence. Finally, we proved the global convergence of E-MS algorithm. |
| |
| Keywords: | missing data model selection observed GIC E-MS algorithm |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《吉林大学学报(理学版)》浏览原始摘要信息 |
| 点击此处可从《吉林大学学报(理学版)》下载免费的PDF全文 |
|
