Your Policy Regularizer is Secretly an Adversary

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Brekelmans, Rob, Genewein, Tim, Grau-Moya, Jordi, Delétang, Grégoire, Kunesch, Markus, Legg, Shane, Ortega, Pedro
Format: Preprint
Veröffentlicht: 2022
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909182126456832
author Brekelmans, Rob
Genewein, Tim
Grau-Moya, Jordi
Delétang, Grégoire
Kunesch, Markus
Legg, Shane
Ortega, Pedro
author_facet Brekelmans, Rob
Genewein, Tim
Grau-Moya, Jordi
Delétang, Grégoire
Kunesch, Markus
Legg, Shane
Ortega, Pedro
contents Policy regularization methods such as maximum entropy regularization are widely used in reinforcement learning to improve the robustness of a learned policy. In this paper, we show how this robustness arises from hedging against worst-case perturbations of the reward function, which are chosen from a limited set by an imagined adversary. Using convex duality, we characterize this robust set of adversarial reward perturbations under KL and alpha-divergence regularization, which includes Shannon and Tsallis entropy regularization as special cases. Importantly, generalization guarantees can be given within this robust set. We provide detailed discussion of the worst-case reward perturbations, and present intuitive empirical examples to illustrate this robustness and its relationship with generalization. Finally, we discuss how our analysis complements and extends previous results on adversarial reward robustness and path consistency optimality conditions.
format Preprint
id arxiv_https___arxiv_org_abs_2203_12592
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Your Policy Regularizer is Secretly an Adversary
Brekelmans, Rob
Genewein, Tim
Grau-Moya, Jordi
Delétang, Grégoire
Kunesch, Markus
Legg, Shane
Ortega, Pedro
Machine Learning
Policy regularization methods such as maximum entropy regularization are widely used in reinforcement learning to improve the robustness of a learned policy. In this paper, we show how this robustness arises from hedging against worst-case perturbations of the reward function, which are chosen from a limited set by an imagined adversary. Using convex duality, we characterize this robust set of adversarial reward perturbations under KL and alpha-divergence regularization, which includes Shannon and Tsallis entropy regularization as special cases. Importantly, generalization guarantees can be given within this robust set. We provide detailed discussion of the worst-case reward perturbations, and present intuitive empirical examples to illustrate this robustness and its relationship with generalization. Finally, we discuss how our analysis complements and extends previous results on adversarial reward robustness and path consistency optimality conditions.
title Your Policy Regularizer is Secretly an Adversary
topic Machine Learning
url https://arxiv.org/abs/2203.12592