ICML 2018oral65 citations

Regret Minimization for Partially Observable Deep Reinforcement Learning

Peter Jin, Kurt Keutzer, Sergey Levine

Abstract

Deep reinforcement learning algorithms that estimate state and state-action value functions have been shown to be effective in a variety of challenging domains, including learning control strategies from raw image pixels. However, algorithms that estimate state and state-action value functions typically assume a fully observed state and must compensate for partial observations by using finite length observation histories or recurrent networks. In this work, we propose a new deep reinforcement learning algorithm based on counterfactual regret minimization that iteratively updates an approximation to an advantage-like function and is robust to partially observed state. We demonstrate that this new algorithm can substantially outperform strong baseline methods on several partially observed reinforcement learning tasks: learning first-person 3D navigation in Doom and Minecraft, and acting in the presence of partially observed objects in Doom and Pong.

BibTeX
@InProceedings{pmlr-v80-jin18c,
  title = 	 {Regret Minimization for Partially Observable Deep Reinforcement Learning},
  author =       {Jin, Peter and Keutzer, Kurt and Levine, Sergey},
  booktitle = 	 {Proceedings of the 35th International Conference on Machine Learning},
  pages = 	 {2342--2351},
  year = 	 {2018},
  editor = 	 {Dy, Jennifer and Krause, Andreas},
  volume = 	 {80},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {10--15 Jul},
  publisher =    {PMLR},
  pdf = 	 {http://proceedings.mlr.press/v80/jin18c/jin18c.pdf},
  url = 	 {https://proceedings.mlr.press/v80/jin18c.html},
  abstract = 	 {Deep reinforcement learning algorithms that estimate state and state-action value functions have been shown to be effective in a variety of challenging domains, including learning control strategies from raw image pixels. However, algorithms that estimate state and state-action value functions typically assume a fully observed state and must compensate for partial observations by using finite length observation histories or recurrent networks. In this work, we propose a new deep reinforcement learning algorithm based on counterfactual regret minimization that iteratively updates an approximation to an advantage-like function and is robust to partially observed state. We demonstrate that this new algorithm can substantially outperform strong baseline methods on several partially observed reinforcement learning tasks: learning first-person 3D navigation in Doom and Minecraft, and acting in the presence of partially observed objects in Doom and Pong.}
}
Regret Minimization for Partially Observable Deep Reinforcement Learning · ICML 2018