High Confidence Off-Policy Evaluation. Thomas, P. S.; Theocharous, G.; and Ghavamzadeh, M. In Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence, of AAAI'15, pages 3000–3006. AAAI Press.
High Confidence Off-Policy Evaluation [link]Paper  abstract   bibtex   
Many reinforcement learning algorithms use trajectories collected from the execution of one or more policies to propose a new policy. Because execution of a bad policy can be costly or dangerous, techniques for evaluating the performance of the new policy without requiring its execution have been of recent interest in industry. Such off-policy evaluation methods, which estimate the performance of a policy using trajectories collected from the execution of other policies, heretofore have not provided confidences regarding the accuracy of their estimates. In this paper we propose an off-policy method for computing a lower confidence bound on the expected return of a policy.
@inproceedings{thomasHighConfidenceOffpolicy2015,
  title = {High {{Confidence Off}}-Policy {{Evaluation}}},
  isbn = {978-0-262-51129-2},
  url = {https://www.aaai.org/ocs/index.php/AAAI/AAAI15/paper/viewPaper/10042},
  abstract = {Many reinforcement learning algorithms use trajectories collected from the execution of one or more policies to propose a new policy. Because execution of a bad policy can be costly or dangerous, techniques for evaluating the performance of the new policy without requiring its execution have been of recent interest in industry. Such off-policy evaluation methods, which estimate the performance of a policy using trajectories collected from the execution of other policies, heretofore have not provided confidences regarding the accuracy of their estimates. In this paper we propose an off-policy method for computing a lower confidence bound on the expected return of a policy.},
  booktitle = {Proceedings of the {{Twenty}}-{{Ninth AAAI Conference}} on {{Artificial Intelligence}}},
  series = {{{AAAI}}'15},
  publisher = {{AAAI Press}},
  urldate = {2019-05-17},
  date = {2015},
  pages = {3000--3006},
  author = {Thomas, Philip S. and Theocharous, Georgios and Ghavamzadeh, Mohammad},
  file = {/home/dimitri/Nextcloud/Zotero/storage/DSD86HE9/Thomas et al. - 2015 - High Confidence Off-policy Evaluation.pdf},
  venue = {Austin, Texas}
}
Downloads: 0