Efficient Gradient Computation for Conditional Gaussian Models
Abstract
We introduce Recursive Exponential Mixed Models (REMMs) and derive the gradient of the parameters for the incomplete-data likelihood. We demonstrate how one can use probabilistic inference in Conditional Gaussian (CG) graphical models, a special case of REMMs, to compute the gradient for a CG model. We also demonstrate that this approach can yield simple and effective algorithms for computing the gradient for models with tied parameters and illustrate this approach on stochastic ARMA models.
Cite
Text
Thiesson and Meek. "Efficient Gradient Computation for Conditional Gaussian Models." Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics, 2005.Markdown
[Thiesson and Meek. "Efficient Gradient Computation for Conditional Gaussian Models." Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics, 2005.](https://mlanthology.org/aistats/2005/thiesson2005aistats-efficient/)BibTeX
@inproceedings{thiesson2005aistats-efficient,
title = {{Efficient Gradient Computation for Conditional Gaussian Models}},
author = {Thiesson, Bo and Meek, Chris},
booktitle = {Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics},
year = {2005},
pages = {341-348},
volume = {R5},
url = {https://mlanthology.org/aistats/2005/thiesson2005aistats-efficient/}
}