Bounding the Partition Function Using Holder's Inequality
Abstract
We describe an algorithm for approximate inference in graphical models based on Holder's inequality that provides upper and lower bounds on common summation problems such as computing the partition function or probability of evidence in a graphical model. Our algorithm unifies and extends several existing approaches, including variable elimination techniques such as mini-bucket elimination and variational methods such as tree reweighted belief propagation and conditional entropy decomposition. We show that our method inherits benefits from each approach to provide significantly better bounds on sum-product tasks.
Cite
Text
Liu and Ihler. "Bounding the Partition Function Using Holder's Inequality." International Conference on Machine Learning, 2011.Markdown
[Liu and Ihler. "Bounding the Partition Function Using Holder's Inequality." International Conference on Machine Learning, 2011.](https://mlanthology.org/icml/2011/liu2011icml-bounding/)BibTeX
@inproceedings{liu2011icml-bounding,
title = {{Bounding the Partition Function Using Holder's Inequality}},
author = {Liu, Qiang and Ihler, Alexander},
booktitle = {International Conference on Machine Learning},
year = {2011},
pages = {849-856},
url = {https://mlanthology.org/icml/2011/liu2011icml-bounding/}
}