Corpus ID: 220633409

Discovering Reinforcement Learning Algorithms

@article{Oh2020DiscoveringRL,
  title={Discovering Reinforcement Learning Algorithms},
  author={Junhyuk Oh and Matteo Hessel and W. Czarnecki and Zhongwen Xu and H. V. Hasselt and S. Singh and D. Silver},
  journal={ArXiv},
  year={2020},
  volume={abs/2007.08794}
}
  • Junhyuk Oh, Matteo Hessel, +4 authors D. Silver
  • Published 2020
  • Computer Science
  • ArXiv
  • Reinforcement learning (RL) algorithms update an agent’s parameters according to one of several possible rules, discovered manually through years of research. Automating the discovery of update rules from data could lead to more efficient algorithms, or algorithms that are better adapted to specific environments. Although there have been prior attempts at addressing this significant scientific challenge, it remains an open question whether it is feasible to discover alternatives to fundamental… CONTINUE READING

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 31 REFERENCES
    Meta-Gradient Reinforcement Learning
    94
    Evolved Policy Gradients
    77
    On Learning Intrinsic Rewards for Policy Gradient Methods
    46
    Behaviour Suite for Reinforcement Learning
    28
    Reinforcement Learning: An Introduction
    25115
    On First-Order Meta-Learning Algorithms
    410