Abstract
Reinforcement learning has had great empirical success in different domains, which has left theoretical foundations, such as performance guarantees, lagging behind. The usual asymptotic convergence to an optimal policy is not strong enough for applications in the real world. Meta learning algorithms aim to use experience from multiple tasks to increase performance on all tasks individually and decrease time taken to reach an acceptable policy. This paper proposes to study the provable properties of meta-reinforcement learning.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Al-Shedivat, M., Bansal, T., Burda, Y., Sutskever, I., Mordatch, I., Abbeel, P.: Continuous adaptation via meta-learning in nonstationary and competitive environments, pp. 1–21, March 2017
Aziz, M., Anderton, J., Kaufmann, E., Aslam, J.: Pure exploration in infinitely-armed bandit models with fixed-confidence, pp. 1–22 (2018)
Brunskill, E.: PAC continuous state online multitask reinforcement learning with identification. In: AAMAS 2016, pp. 438–446 (2016)
Finn, C., Abbeel, P., Levine, S.: Model-agnostic meta-learning for fast adaptation of deep networks (2017)
Jiang, N., Krishnamurthy, A., Agarwal, A., Langford, J., Schapire, R.E.: Contextual decision processes with low Bellman rank are PAC-learnable, pp. 1–42 (2016)
Mnih, V., et al.: Human-level control through deep reinforcement learning. Nature 518, 529 (2015)
Silver, D., et al.: Mastering the game of Go with deep neural networks and tree search. Nature 529, 484 (2016)
Sutton, R.S., Barto, A.G.: Reinforcement learning: an introduction. UCL, Computer Science Department, Reinforcement Learning Lectures, p. 1054 (2017)
Wang, J.X., et al.: Prefrontal cortex as a meta-reinforcement learning system. Nat. Neurosci. 21(6), 860–868 (2018)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer Nature Switzerland AG
About this paper
Cite this paper
Mahony, A. (2018). Formalising Performance Guarantees in Meta-Reinforcement Learning. In: Sun, J., Sun, M. (eds) Formal Methods and Software Engineering. ICFEM 2018. Lecture Notes in Computer Science(), vol 11232. Springer, Cham. https://doi.org/10.1007/978-3-030-02450-5_37
Download citation
DOI: https://doi.org/10.1007/978-3-030-02450-5_37
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-02449-9
Online ISBN: 978-3-030-02450-5
eBook Packages: Computer ScienceComputer Science (R0)