Learning Predictive Representations from a History

Abstract

Predictive State Representations (PSRs) have shown a great deal of promise as an alternative to Markov models. However, learning a PSR from a single stream of data generated from an environment remains a challenge. In this work, we present a formalism of PSRs and the domains they model. This formalization suggests an algorithm for learning PSRs that will (almost surely) converge to a globally optimal model given sufficient training data.

Cite

Text

Wiewiora. "Learning Predictive Representations from a History." International Conference on Machine Learning, 2005. doi:10.1145/1102351.1102473

Markdown

[Wiewiora. "Learning Predictive Representations from a History." International Conference on Machine Learning, 2005.](https://mlanthology.org/icml/2005/wiewiora2005icml-learning/) doi:10.1145/1102351.1102473

BibTeX

@inproceedings{wiewiora2005icml-learning,
  title     = {{Learning Predictive Representations from a History}},
  author    = {Wiewiora, Eric},
  booktitle = {International Conference on Machine Learning},
  year      = {2005},
  pages     = {964-971},
  doi       = {10.1145/1102351.1102473},
  url       = {https://mlanthology.org/icml/2005/wiewiora2005icml-learning/}
}