DMCA
Gibbs max-margin topic models with data augmentation
| Venue: | Journal of Machine Learning Research (JMLR |
| Citations: | 13 - 5 self |
BibTeX
@ARTICLE{Zhu_gibbsmax-margin,
author = {Jun Zhu and Ning Chen and Hugh Perkins and Bo Zhang and David Blei},
title = {Gibbs max-margin topic models with data augmentation},
journal = {Journal of Machine Learning Research (JMLR},
year = {},
pages = {2014}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Max-margin learning is a powerful approach to building classifiers and structured output predictors. Recent work on max-margin supervised topic models has successfully integrated it with Bayesian topic models to discover discriminative latent semantic structures and make accurate predictions for unseen testing data. However, the resulting learning prob-lems are usually hard to solve because of the non-smoothness of the margin loss. Existing approaches to building max-margin supervised topic models rely on an iterative procedure to solve multiple latent SVM subproblems with additional mean-field assumptions on the desired posterior distributions. This paper presents an alternative approach by defining a new max-margin loss. Namely, we present Gibbs max-margin supervised topic models, a latent variable Gibbs classifier to discover hidden topic representations for various tasks, including classification, regression and multi-task learning. Gibbs max-margin supervised topic models minimize an expected margin loss, which is an upper bound of the existing margin loss derived from an expected prediction rule. By introducing augmented variables and integrating out the Dirichlet variables analytically by conjugacy, we develop simple
Keyphrases
topic model gibbs max-margin topic model data augmentation gibbs max-margin margin loss learning prob-lems additional mean-field assumption discriminative latent semantic structure posterior distribution recent work new max-margin loss latent variable gibbs classifier output predictor max-margin learning multiple latent svm bayesian topic model prediction rule accurate prediction augmented variable hidden topic representation dirichlet variable powerful approach iterative procedure various task alternative approach multi-task learning upper bound expected margin loss
