Learning the Linear Dynamical System with ASOS
Abstract
We develop a new algorithm, based on EM, for learning the Linear Dynamical System model. Called the method of Approximated Second-Order Statistics (ASOS) our approach achieves dramatically superior computational performance over standard EM through its use of approximations, which we justify with both intuitive explanations and rigorous convergence results. In particular, after an inexpensive pre-computation phase, the iterations of ASOS can be performed in time independent of the length of the training dataset.
Cite
Text
Martens. "Learning the Linear Dynamical System with ASOS." International Conference on Machine Learning, 2010.Markdown
[Martens. "Learning the Linear Dynamical System with ASOS." International Conference on Machine Learning, 2010.](https://mlanthology.org/icml/2010/martens2010icml-learning/)BibTeX
@inproceedings{martens2010icml-learning,
title = {{Learning the Linear Dynamical System with ASOS}},
author = {Martens, James},
booktitle = {International Conference on Machine Learning},
year = {2010},
pages = {743-750},
url = {https://mlanthology.org/icml/2010/martens2010icml-learning/}
}