arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2110.04232stat.MLcs.IRcs.LGstat.ME
收藏

学习主题模型的可识别性与样本分析

Learning Topic Models: Identifiability and Finite-Sample Analysis

Meta Platforms, Inc.(元平台公司);University of Illinois at Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校)

Yinyin Chen, Shishuang He, Yun Yang, Feng Liang

摘要:

主题模型为学习、提取和发现大型文本语料库中的潜在结构提供了一种有用的文本挖掘工具。虽然已经提出了大量的方法为主题建模,缺乏在文献中是一个正式的理论研究的统计识别性和准确性的潜在主题估计。在本文中,我们提出了一个极大似然估计(MLE)的基础上,一个特定的综合可能性,自然连接到的概念,在计算几何,体积最小化的潜在主题。我们的理论引入了一组新的几何条件的主题模型的可识别性,条件是弱于传统的可分性条件,通常依赖于纯主题文档或锚词的存在。较弱的条件允许更广泛的,从而可能更富有成效的调查。我们进行有限样本误差分析,建议的估计,并讨论我们的结果和以前的调查之间的连接。最后,我们采用模拟和真实数据集的实证研究。

英文摘要:

Topic models provide a useful text-mining tool for learning, extracting, and discovering latent structures in large text corpora. Although a plethora of methods have been proposed for topic modeling, lacking in the literature is a formal theoretical investigation of the statistical identifiability and accuracy of latent topic estimation. In this paper, we propose a maximum likelihood estimator (MLE) of latent topics based on a specific integrated likelihood that is naturally connected to the concept, in computational geometry, of volume minimization. Our theory introduces a new set of geometric conditions for topic model identifiability, conditions that are weaker than conventional separability conditions, which typically rely on the existence of pure topic documents or of anchor words. Weaker conditions allow a wider and thus potentially more fruitful investigation. We conduct finite-sample error analysis for the proposed estimator and discuss connections between our results and those of previous investigations. We conclude with empirical studies employing both simulated and real datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑