主题模型评价方法Evaluation Methods for Topic Models |
|
课程网址: | http://videolectures.net/icml09_wallach_emtm/ |
主讲教师: | Hanna M. Wallach |
开课单位: | 马萨诸塞大学 |
开课时间: | 2009-08-26 |
课程语种: | 英语 |
中文简介: | 统计主题模型的自然评估度量是给定训练模型的持有文档的概率。虽然这种概率的精确计算是难以处理的,但是在主题建模文献中已经使用了这种概率的几个估计量,包括调和平均法和经验似然法。在本文中,我们通过实验证明,常用的方法不太可能准确地估计出文件的概率,并提出了两种既准确又有效的替代方法。 |
课程简介: | A natural evaluation metric for statistical topic models is the probability of held-out documents given a trained model. While exact computation of this probability is intractable, several estimators for this probability have been used in the topic modeling literature, including the harmonic mean method and empirical likelihood method. In this paper, we demonstrate experimentally that commonly-used methods are unlikely to accurately estimate the probability of held-out documents, and propose two alternative methods that are both accurate and efficient. |
关 键 词: | 统计主题模型; 主题建模; 调和平均法 |
课程来源: | 视频讲座网 |
最后编审: | 2020-07-13:yumf |
阅读次数: | 212 |