Policy Gradient Methods in the Presence of Symmetries and State Abstractions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Panangaden, Prakash, Rezaei-Shoshtari, Sahand, Zhao, Rosie, Meger, David, Precup, Doina
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914706582667264
author Panangaden, Prakash
Rezaei-Shoshtari, Sahand
Zhao, Rosie
Meger, David
Precup, Doina
author_facet Panangaden, Prakash
Rezaei-Shoshtari, Sahand
Zhao, Rosie
Meger, David
Precup, Doina
contents Reinforcement learning (RL) on high-dimensional and complex problems relies on abstraction for improved efficiency and generalization. In this paper, we study abstraction in the continuous-control setting, and extend the definition of Markov decision process (MDP) homomorphisms to the setting of continuous state and action spaces. We derive a policy gradient theorem on the abstract MDP for both stochastic and deterministic policies. Our policy gradient results allow for leveraging approximate symmetries of the environment for policy optimization. Based on these theorems, we propose a family of actor-critic algorithms that are able to learn the policy and the MDP homomorphism map simultaneously, using the lax bisimulation metric. Finally, we introduce a series of environments with continuous symmetries to further demonstrate the ability of our algorithm for action abstraction in the presence of such symmetries. We demonstrate the effectiveness of our method on our environments, as well as on challenging visual control tasks from the DeepMind Control Suite. Our method's ability to utilize MDP homomorphisms for representation learning leads to improved performance, and the visualizations of the latent space clearly demonstrate the structure of the learned abstraction.
format Preprint
id arxiv_https___arxiv_org_abs_2305_05666
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Policy Gradient Methods in the Presence of Symmetries and State Abstractions
Panangaden, Prakash
Rezaei-Shoshtari, Sahand
Zhao, Rosie
Meger, David
Precup, Doina
Machine Learning
Artificial Intelligence
Reinforcement learning (RL) on high-dimensional and complex problems relies on abstraction for improved efficiency and generalization. In this paper, we study abstraction in the continuous-control setting, and extend the definition of Markov decision process (MDP) homomorphisms to the setting of continuous state and action spaces. We derive a policy gradient theorem on the abstract MDP for both stochastic and deterministic policies. Our policy gradient results allow for leveraging approximate symmetries of the environment for policy optimization. Based on these theorems, we propose a family of actor-critic algorithms that are able to learn the policy and the MDP homomorphism map simultaneously, using the lax bisimulation metric. Finally, we introduce a series of environments with continuous symmetries to further demonstrate the ability of our algorithm for action abstraction in the presence of such symmetries. We demonstrate the effectiveness of our method on our environments, as well as on challenging visual control tasks from the DeepMind Control Suite. Our method's ability to utilize MDP homomorphisms for representation learning leads to improved performance, and the visualizations of the latent space clearly demonstrate the structure of the learned abstraction.
title Policy Gradient Methods in the Presence of Symmetries and State Abstractions
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2305.05666