A Unifying Representation for a Class of Dependent Random Measures
Abstract
We present a general construction for dependent random measures based on thinning Poisson processes on an augmented space. The framework is not restricted to dependent versions of a specific nonparametric model, but can be applied to all models that can be represented using completely random measures. Several existing dependent random measures can be seen as specific cases of this framework. Interesting properties of the resulting measures are derived and the efficacy of the framework is demonstrated by constructing a covariate-dependent latent feature model and topic model that obtain superior predictive performance.
Cite
Text
Foti et al. "A Unifying Representation for a Class of Dependent Random Measures." International Conference on Artificial Intelligence and Statistics, 2013.Markdown
[Foti et al. "A Unifying Representation for a Class of Dependent Random Measures." International Conference on Artificial Intelligence and Statistics, 2013.](https://mlanthology.org/aistats/2013/foti2013aistats-unifying/)BibTeX
@inproceedings{foti2013aistats-unifying,
title = {{A Unifying Representation for a Class of Dependent Random Measures}},
author = {Foti, Nicholas J. and Futoma, Joseph D. and Rockmore, Daniel N. and Williamson, Sinead},
booktitle = {International Conference on Artificial Intelligence and Statistics},
year = {2013},
pages = {20-28},
url = {https://mlanthology.org/aistats/2013/foti2013aistats-unifying/}
}