Sample Efficient Policy Search for Optimal Stopping Domains
Abstract
Optimal stopping problems consider the question of deciding when to stop an observation-generating process in order to maximize a return. We examine the problem of simultaneously learning and planning in such domains, when data is collected directly from the environment. We propose GFSE, a simple and flexible model-free policy search method that reuses data for sample efficiency by leveraging problem structure. We bound the sample complexity of our approach to guarantee uniform convergence of policy value estimates, tightening existing PAC bounds to achieve logarithmic dependence on horizon length for our setting. We also examine the benefit of our method against prevalent model-based and model-free approaches on 3 domains taken from diverse fields.
Cite
Text
Goel et al. "Sample Efficient Policy Search for Optimal Stopping Domains." International Joint Conference on Artificial Intelligence, 2017. doi:10.24963/IJCAI.2017/237Markdown
[Goel et al. "Sample Efficient Policy Search for Optimal Stopping Domains." International Joint Conference on Artificial Intelligence, 2017.](https://mlanthology.org/ijcai/2017/goel2017ijcai-sample/) doi:10.24963/IJCAI.2017/237BibTeX
@inproceedings{goel2017ijcai-sample,
title = {{Sample Efficient Policy Search for Optimal Stopping Domains}},
author = {Goel, Karan and Dann, Christoph and Brunskill, Emma},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {2017},
pages = {1711-1717},
doi = {10.24963/IJCAI.2017/237},
url = {https://mlanthology.org/ijcai/2017/goel2017ijcai-sample/}
}