基于EM算法的Dirichlet分布的参数估计  

The parameter estimate based on EM algorithm for the Dirichlet distribution

作　　者：吴琴[1] 葛文秀 WU Qin;GE Wen-xiu(School of Mathematical Sciences,South China Normal University,Guangzhou 510631,China)

机构地区：[1]华南师范大学数学科学学院,广东广州510631

出　　处：《东北师大学报（自然科学版）》2021年第4期30-33,共4页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(11401226);广东省自然科学基金博士启动基金资助项目(2017A030310264).

摘　　要：为了克服Dirichlet分布的传统估计算法复杂且不能保证有效性的缺陷,将基于Dirichlet分布的随机表示引入缺失数据,构造了EM算法计算参数的极大似然估计,算法简单的同时保证了估计的有效性.最后进行统计模拟,结果表明,本文提出的基于EM算法的Dirichlet分布的参数估计有很好的估计精度.To overcome the complicity and ineffectiveness in the traditional algorithm in the parameter estimation for the Dirichlet distribution,the stochastic representation for the Dirichlet distribution is proposed.Based on the stochastic representation,the missing data are introduced and EM algorithm is employed to calculate the MLE.The algorithm is simple and effective.The simulation results showed the estimate based on EM algorithm is accurate.

关 键 词：Dirichlet分布 EM算法 GAMMA分布 极大似然估计 

分 类 号：O212[理学—概率论与数理统计]