arXiv Analytics

Sign in

arXiv:1809.01560 [cs.LG]AbstractReferencesReviewsResources

Reinforcement Learning under Threats

Víctor Gallego, Roi Naveiro, David Ríos Insua

Published 2018-09-05Version 1

In several reinforcement learning (RL) scenarios, mainly in security settings, there may be adversaries trying to interfere with the reward generating process. In this paper, we introduce Threatened Markov Decision Processes (TMDPs), which provide a framework to support a decision maker against a potential adversary in RL. Furthermore, we propose a level-$k$ thinking scheme resulting in a new learning framework to deal with TMDPs. After introducing our framework and deriving theoretical results, relevant empirical evidence is given via extensive experiments, showing the benefits of accounting for adversaries while the agent learns.

Related articles: Most relevant | Search more
arXiv:1809.10679 [cs.LG] (Published 2018-09-27)
Definition and evaluation of model-free coordination of electrical vehicle charging with reinforcement learning
arXiv:1803.00590 [cs.LG] (Published 2018-03-01)
Hierarchical Imitation and Reinforcement Learning
arXiv:1809.09095 [cs.LG] (Published 2018-09-23)
On Reinforcement Learning for Full-length Game of StarCraft