Fingerprint Policy Optimisation for Robust Reinforcement Learning

Supratik Paul, Michael A. Osborne, Shimon Whiteson
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:5082-5091, 2019.

Abstract

Policy gradient methods ignore the potential value of adjusting environment variables: unobservable state features that are randomly determined by the environment in a physical setting, but are controllable in a simulator. This can lead to slow learning, or convergence to suboptimal policies, if the environment variable has a large impact on the transition dynamics. In this paper, we present fingerprint policy optimisation (FPO), which finds a policy that is optimal in expectation across the distribution of environment variables. The central idea is to use Bayesian optimisation (BO) to actively select the distribution of the environment variable that maximises the improvement generated by each iteration of the policy gradient method. To make this BO practical, we contribute two easy-to-compute low-dimensional fingerprints of the current policy. Our experiments show that FPO can efficiently learn policies that are robust to significant rare events, which are unlikely to be observable under random sampling, but are key to learning good policies.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-paul19a, title = {Fingerprint Policy Optimisation for Robust Reinforcement Learning}, author = {Paul, Supratik and Osborne, Michael A. and Whiteson, Shimon}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {5082--5091}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/paul19a/paul19a.pdf}, url = {https://proceedings.mlr.press/v97/paul19a.html}, abstract = {Policy gradient methods ignore the potential value of adjusting environment variables: unobservable state features that are randomly determined by the environment in a physical setting, but are controllable in a simulator. This can lead to slow learning, or convergence to suboptimal policies, if the environment variable has a large impact on the transition dynamics. In this paper, we present fingerprint policy optimisation (FPO), which finds a policy that is optimal in expectation across the distribution of environment variables. The central idea is to use Bayesian optimisation (BO) to actively select the distribution of the environment variable that maximises the improvement generated by each iteration of the policy gradient method. To make this BO practical, we contribute two easy-to-compute low-dimensional fingerprints of the current policy. Our experiments show that FPO can efficiently learn policies that are robust to significant rare events, which are unlikely to be observable under random sampling, but are key to learning good policies.} }
Endnote
%0 Conference Paper %T Fingerprint Policy Optimisation for Robust Reinforcement Learning %A Supratik Paul %A Michael A. Osborne %A Shimon Whiteson %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-paul19a %I PMLR %P 5082--5091 %U https://proceedings.mlr.press/v97/paul19a.html %V 97 %X Policy gradient methods ignore the potential value of adjusting environment variables: unobservable state features that are randomly determined by the environment in a physical setting, but are controllable in a simulator. This can lead to slow learning, or convergence to suboptimal policies, if the environment variable has a large impact on the transition dynamics. In this paper, we present fingerprint policy optimisation (FPO), which finds a policy that is optimal in expectation across the distribution of environment variables. The central idea is to use Bayesian optimisation (BO) to actively select the distribution of the environment variable that maximises the improvement generated by each iteration of the policy gradient method. To make this BO practical, we contribute two easy-to-compute low-dimensional fingerprints of the current policy. Our experiments show that FPO can efficiently learn policies that are robust to significant rare events, which are unlikely to be observable under random sampling, but are key to learning good policies.
APA
Paul, S., Osborne, M.A. & Whiteson, S.. (2019). Fingerprint Policy Optimisation for Robust Reinforcement Learning. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:5082-5091 Available from https://proceedings.mlr.press/v97/paul19a.html.

Related Material