Tractable learning and inference with high-order representations
A. Culotta, and A. Mccallum. In ICML Workshop on Open Problems in Statistical Relational Learning, (2006)
Abstract
Representing high-order interactions in data often results in large models with an intractable number of hidden variables. In these models, inference and learning must operate without instantiating the entire set of variables. This paper presents a Metropolis-Hastings sampling approach to address this issue, and proposes new methods to discriminatively estimate the proposal and target distribution of the sampler using a ranking function over configurations. We demonstrate our approach on the task of paper and author deduplication, showing that our method enables complex, advantageous representations of the data while maintaining tractable learning and inference procedures. 1.
Description
CiteSeerX — Tractable learning and inference with high-order representations
%0 Conference Paper
%1 culotta06tractable
%A Culotta, Aron
%A Mccallum, Andrew
%B In ICML Workshop on Open Problems in Statistical Relational Learning
%D 2006
%K MAP MarkovLogic SRL inference learning
%T Tractable learning and inference with high-order representations
%U http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.117.3785
%X Representing high-order interactions in data often results in large models with an intractable number of hidden variables. In these models, inference and learning must operate without instantiating the entire set of variables. This paper presents a Metropolis-Hastings sampling approach to address this issue, and proposes new methods to discriminatively estimate the proposal and target distribution of the sampler using a ranking function over configurations. We demonstrate our approach on the task of paper and author deduplication, showing that our method enables complex, advantageous representations of the data while maintaining tractable learning and inference procedures. 1.
@inproceedings{culotta06tractable,
abstract = {Representing high-order interactions in data often results in large models with an intractable number of hidden variables. In these models, inference and learning must operate without instantiating the entire set of variables. This paper presents a Metropolis-Hastings sampling approach to address this issue, and proposes new methods to discriminatively estimate the proposal and target distribution of the sampler using a ranking function over configurations. We demonstrate our approach on the task of paper and author deduplication, showing that our method enables complex, advantageous representations of the data while maintaining tractable learning and inference procedures. 1.},
added-at = {2010-10-27T16:23:40.000+0200},
author = {Culotta, Aron and Mccallum, Andrew},
biburl = {https://www.bibsonomy.org/bibtex/25eb36972bda6115fa0fd7a8f2516a831/djain},
booktitle = {In ICML Workshop on Open Problems in Statistical Relational Learning},
description = {CiteSeerX — Tractable learning and inference with high-order representations},
interhash = {a0947a3c7e20ddd14bb6eb159524030e},
intrahash = {5eb36972bda6115fa0fd7a8f2516a831},
keywords = {MAP MarkovLogic SRL inference learning},
timestamp = {2010-10-27T16:24:11.000+0200},
title = {Tractable learning and inference with high-order representations},
url = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.117.3785},
year = 2006
}