FeUdal Networks for Hierarchical Reinforcement Learning

Alexander Sasha Vezhnevets, Simon Osindero, Tom Schaul, Nicolas Heess, Max Jaderberg, David Silver, Koray Kavukcuoglu
Proceedings of the 34th International Conference on Machine Learning, PMLR 70:3540-3549, 2017.

Abstract

We introduce FeUdal Networks (FuNs): a novel architecture for hierarchical reinforcement learning. Our approach is inspired by the feudal reinforcement learning proposal of Dayan and Hinton, and gains power and efficacy by decoupling end-to-end learning across multiple levels – allowing it to utilise different resolutions of time. Our framework employs a Manager module and a Worker module. The Manager operates at a slower time scale and sets abstract goals which are conveyed to and enacted by the Worker. The Worker generates primitive actions at every tick of the environment. The decoupled structure of FuN conveys several benefits – in addition to facilitating very long timescale credit assignment it also encourages the emergence of sub-policies associated with different goals set by the Manager. These properties allow FuN to dramatically outperform a strong baseline agent on tasks that involve long-term credit assignment or memorisation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v70-vezhnevets17a, title = {{F}e{U}dal Networks for Hierarchical Reinforcement Learning}, author = {Alexander Sasha Vezhnevets and Simon Osindero and Tom Schaul and Nicolas Heess and Max Jaderberg and David Silver and Koray Kavukcuoglu}, booktitle = {Proceedings of the 34th International Conference on Machine Learning}, pages = {3540--3549}, year = {2017}, editor = {Precup, Doina and Teh, Yee Whye}, volume = {70}, series = {Proceedings of Machine Learning Research}, month = {06--11 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v70/vezhnevets17a/vezhnevets17a.pdf}, url = {https://proceedings.mlr.press/v70/vezhnevets17a.html}, abstract = {We introduce FeUdal Networks (FuNs): a novel architecture for hierarchical reinforcement learning. Our approach is inspired by the feudal reinforcement learning proposal of Dayan and Hinton, and gains power and efficacy by decoupling end-to-end learning across multiple levels – allowing it to utilise different resolutions of time. Our framework employs a Manager module and a Worker module. The Manager operates at a slower time scale and sets abstract goals which are conveyed to and enacted by the Worker. The Worker generates primitive actions at every tick of the environment. The decoupled structure of FuN conveys several benefits – in addition to facilitating very long timescale credit assignment it also encourages the emergence of sub-policies associated with different goals set by the Manager. These properties allow FuN to dramatically outperform a strong baseline agent on tasks that involve long-term credit assignment or memorisation.} }
Endnote
%0 Conference Paper %T FeUdal Networks for Hierarchical Reinforcement Learning %A Alexander Sasha Vezhnevets %A Simon Osindero %A Tom Schaul %A Nicolas Heess %A Max Jaderberg %A David Silver %A Koray Kavukcuoglu %B Proceedings of the 34th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2017 %E Doina Precup %E Yee Whye Teh %F pmlr-v70-vezhnevets17a %I PMLR %P 3540--3549 %U https://proceedings.mlr.press/v70/vezhnevets17a.html %V 70 %X We introduce FeUdal Networks (FuNs): a novel architecture for hierarchical reinforcement learning. Our approach is inspired by the feudal reinforcement learning proposal of Dayan and Hinton, and gains power and efficacy by decoupling end-to-end learning across multiple levels – allowing it to utilise different resolutions of time. Our framework employs a Manager module and a Worker module. The Manager operates at a slower time scale and sets abstract goals which are conveyed to and enacted by the Worker. The Worker generates primitive actions at every tick of the environment. The decoupled structure of FuN conveys several benefits – in addition to facilitating very long timescale credit assignment it also encourages the emergence of sub-policies associated with different goals set by the Manager. These properties allow FuN to dramatically outperform a strong baseline agent on tasks that involve long-term credit assignment or memorisation.
APA
Vezhnevets, A.S., Osindero, S., Schaul, T., Heess, N., Jaderberg, M., Silver, D. & Kavukcuoglu, K.. (2017). FeUdal Networks for Hierarchical Reinforcement Learning. Proceedings of the 34th International Conference on Machine Learning, in Proceedings of Machine Learning Research 70:3540-3549 Available from https://proceedings.mlr.press/v70/vezhnevets17a.html.

Related Material