Saved in:
Bibliographic Details
Main Authors: Khattar, Vanshaj, Ding, Yuhao, Sel, Bilgehan, Lavaei, Javad, Jin, Ming
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.16601
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911888341729280
author Khattar, Vanshaj
Ding, Yuhao
Sel, Bilgehan
Lavaei, Javad
Jin, Ming
author_facet Khattar, Vanshaj
Ding, Yuhao
Sel, Bilgehan
Lavaei, Javad
Jin, Ming
contents Meta-reinforcement learning has widely been used as a learning-to-learn framework to solve unseen tasks with limited experience. However, the aspect of constraint violations has not been adequately addressed in the existing works, making their application restricted in real-world settings. In this paper, we study the problem of meta-safe reinforcement learning (Meta-SRL) through the CMDP-within-online framework to establish the first provable guarantees in this important setting. We obtain task-averaged regret bounds for the reward maximization (optimality gap) and constraint violations using gradient-based meta-learning and show that the task-averaged optimality gap and constraint satisfaction improve with task-similarity in a static environment or task-relatedness in a dynamic environment. Several technical challenges arise when making this framework practical. To this end, we propose a meta-algorithm that performs inexact online learning on the upper bounds of within-task optimality gap and constraint violations estimated by off-policy stationary distribution corrections. Furthermore, we enable the learning rates to be adapted for every task and extend our approach to settings with a competing dynamically changing oracle. Finally, experiments are conducted to demonstrate the effectiveness of our approach.
format Preprint
id arxiv_https___arxiv_org_abs_2405_16601
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle A CMDP-within-online framework for Meta-Safe Reinforcement Learning
Khattar, Vanshaj
Ding, Yuhao
Sel, Bilgehan
Lavaei, Javad
Jin, Ming
Machine Learning
Meta-reinforcement learning has widely been used as a learning-to-learn framework to solve unseen tasks with limited experience. However, the aspect of constraint violations has not been adequately addressed in the existing works, making their application restricted in real-world settings. In this paper, we study the problem of meta-safe reinforcement learning (Meta-SRL) through the CMDP-within-online framework to establish the first provable guarantees in this important setting. We obtain task-averaged regret bounds for the reward maximization (optimality gap) and constraint violations using gradient-based meta-learning and show that the task-averaged optimality gap and constraint satisfaction improve with task-similarity in a static environment or task-relatedness in a dynamic environment. Several technical challenges arise when making this framework practical. To this end, we propose a meta-algorithm that performs inexact online learning on the upper bounds of within-task optimality gap and constraint violations estimated by off-policy stationary distribution corrections. Furthermore, we enable the learning rates to be adapted for every task and extend our approach to settings with a competing dynamically changing oracle. Finally, experiments are conducted to demonstrate the effectiveness of our approach.
title A CMDP-within-online framework for Meta-Safe Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2405.16601