Improving PAC exploration using the median of means

We present the first application of the median of means in a PAC exploration algorithm for MDPs. Using the median of means allows us to significantly reduce the dependence of our bounds on the range of values that the value function can take, while introducing a dependence on the (potentially much s...

ver descrição completa

Detalhes bibliográficos
Principais autores: Pazis, Jason, How, Jonathan P
Outros Autores: Massachusetts Institute of Technology. Aerospace Controls Laboratory
Formato: Artigo
Publicado em: Neural Information Processing Systems Foundation 2018
Acesso em linha:http://hdl.handle.net/1721.1/114290
https://orcid.org/0000-0001-8576-1930
_version_ 1826210615616602112
author Pazis, Jason
How, Jonathan P
author2 Massachusetts Institute of Technology. Aerospace Controls Laboratory
author_facet Massachusetts Institute of Technology. Aerospace Controls Laboratory
Pazis, Jason
How, Jonathan P
author_sort Pazis, Jason
collection MIT
description We present the first application of the median of means in a PAC exploration algorithm for MDPs. Using the median of means allows us to significantly reduce the dependence of our bounds on the range of values that the value function can take, while introducing a dependence on the (potentially much smaller) variance of the Bellman operator. Additionally, our algorithm is the first algorithm with PAC bounds that can be applied to MDPs with unbounded rewards.
first_indexed 2024-09-23T14:52:45Z
format Article
id mit-1721.1/114290
institution Massachusetts Institute of Technology
last_indexed 2024-09-23T14:52:45Z
publishDate 2018
publisher Neural Information Processing Systems Foundation
record_format dspace
spelling mit-1721.1/1142902022-09-29T11:12:00Z Improving PAC exploration using the median of means Pazis, Jason How, Jonathan P Massachusetts Institute of Technology. Aerospace Controls Laboratory Massachusetts Institute of Technology. Department of Aeronautics and Astronautics Massachusetts Institute of Technology. Laboratory for Information and Decision Systems Pazis, Jason How, Jonathan P We present the first application of the median of means in a PAC exploration algorithm for MDPs. Using the median of means allows us to significantly reduce the dependence of our bounds on the range of values that the value function can take, while introducing a dependence on the (potentially much smaller) variance of the Bellman operator. Additionally, our algorithm is the first algorithm with PAC bounds that can be applied to MDPs with unbounded rewards. United States. Office of Naval Research (Grant N000141110688) National Science Foundation (U.S.) (Grant IIS-1218931) 2018-03-26T14:30:15Z 2018-03-26T14:30:15Z 2016 2018-03-21T18:03:04Z Article http://purl.org/eprint/type/ConferencePaper 1049-5258 http://hdl.handle.net/1721.1/114290 Pazis, Jason et al. "Improving PAC Exploration Using the Median Of Means." Advances in Neural Information Processing Systems (NIPS 2016), 29 (2016) © 2016 NIPS Foundation - All Rights Reserved. https://orcid.org/0000-0001-8576-1930 https://papers.nips.cc/paper/6577-improving-pac-exploration-using-the-median-of-means Advances in Neural Information Processing Systems (NIPS) Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/pdf Neural Information Processing Systems Foundation Neural Information Processing Systems (NIPS)
spellingShingle Pazis, Jason
How, Jonathan P
Improving PAC exploration using the median of means
title Improving PAC exploration using the median of means
title_full Improving PAC exploration using the median of means
title_fullStr Improving PAC exploration using the median of means
title_full_unstemmed Improving PAC exploration using the median of means
title_short Improving PAC exploration using the median of means
title_sort improving pac exploration using the median of means
url http://hdl.handle.net/1721.1/114290
https://orcid.org/0000-0001-8576-1930
work_keys_str_mv AT pazisjason improvingpacexplorationusingthemedianofmeans
AT howjonathanp improvingpacexplorationusingthemedianofmeans