Learning Low Dimensional Predictive Representations
Abstract
Predictive state representations (PSRs) have recently been proposed as an alternative to partially observable Markov decision processes (POMDPs) for representing the state of a dynamical system (Littman et al., 2001). We present a learning algorithm that learns a PSR from observational data. Our algorithm produces a variant of PSRs called transformed predictive state representations (TPSRs). We provide an efficient principal-components-based algorithm for learning a TPSR, and show that TPSRs can perform well in comparison to Hidden Markov Models learned with Baum-Welch in a real world robot tracking task for low dimensional representations and long prediction horizons.
Cite
Text
Rosencrantz et al. "Learning Low Dimensional Predictive Representations." International Conference on Machine Learning, 2004. doi:10.1145/1015330.1015441Markdown
[Rosencrantz et al. "Learning Low Dimensional Predictive Representations." International Conference on Machine Learning, 2004.](https://mlanthology.org/icml/2004/rosencrantz2004icml-learning/) doi:10.1145/1015330.1015441BibTeX
@inproceedings{rosencrantz2004icml-learning,
title = {{Learning Low Dimensional Predictive Representations}},
author = {Rosencrantz, Matthew and Gordon, Geoffrey J. and Thrun, Sebastian},
booktitle = {International Conference on Machine Learning},
year = {2004},
doi = {10.1145/1015330.1015441},
url = {https://mlanthology.org/icml/2004/rosencrantz2004icml-learning/}
}