Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: Combinatorial Topic Models using Small-Variance Asymptotics
(Submitted on 7 Apr 2016 (this version), latest version 27 May 2016 (v2))
Abstract: Topic models have emerged as fundamental tools in unsupervised machine learning. Most modern topic modeling algorithms take a probabilistic view and derive inference algorithms based on Latent Dirichlet Allocation (LDA) or its variants. In contrast, we study topic modeling as a combinatorial optimization problem, and derive its objective function from LDA by passing to the small-variance limit. We minimize the derived objective by using ideas from combinatorial optimization, which results in a new, fast, and high-quality topic modeling algorithm. In particular, we show the surprising result that our algorithm can outperform all major LDA-based topic modeling approaches, even when the data are sampled from an LDA model and true hyper-parameters are provided to these competitors. These results make a strong case that topic models need not be limited to a probabilistic view.
Submission history
From: Ke Jiang [view email][v1] Thu, 7 Apr 2016 15:04:16 GMT (986kb,D)
[v2] Fri, 27 May 2016 03:11:02 GMT (379kb,D)
Link back to: arXiv, form interface, contact.