On Combining Graph-Based Variance Reduction Schemes
Abstract
In this paper, we consider two variance reduction schemes that exploit the structure of the primal graph of the graphical model: Rao-Blackwellised w-cutset sampling and AND/OR sampling. We show that the two schemes are orthogonal and can be combined to further reduce the variance. Our combination yields a new family of estimators which trade time and space with variance. We demonstrate experimentally that the new estimators are superior, often yielding an order of magnitude improvement over previous schemes on several benchmarks.
Cite
Text
Gogate and Dechter. "On Combining Graph-Based Variance Reduction Schemes." Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, 2010.Markdown
[Gogate and Dechter. "On Combining Graph-Based Variance Reduction Schemes." Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, 2010.](https://mlanthology.org/aistats/2010/gogate2010aistats-combining/)BibTeX
@inproceedings{gogate2010aistats-combining,
title = {{On Combining Graph-Based Variance Reduction Schemes}},
author = {Gogate, Vibhav and Dechter, Rina},
booktitle = {Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics},
year = {2010},
pages = {257-264},
volume = {9},
url = {https://mlanthology.org/aistats/2010/gogate2010aistats-combining/}
}