A PAC Learning Algorithm for LTL and Omega-regular Objectives in MDPs

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Perez, Mateo, Somenzi, Fabio, Trivedi, Ashutosh
Formato: Preprint
Publicado: 2023
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914687380094976
author Perez, Mateo
Somenzi, Fabio
Trivedi, Ashutosh
author_facet Perez, Mateo
Somenzi, Fabio
Trivedi, Ashutosh
contents Linear temporal logic (LTL) and omega-regular objectives -- a superset of LTL -- have seen recent use as a way to express non-Markovian objectives in reinforcement learning. We introduce a model-based probably approximately correct (PAC) learning algorithm for omega-regular objectives in Markov decision processes (MDPs). As part of the development of our algorithm, we introduce the epsilon-recurrence time: a measure of the speed at which a policy converges to the satisfaction of the omega-regular objective in the limit. We prove that our algorithm only requires a polynomial number of samples in the relevant parameters, and perform experiments which confirm our theory.
format Preprint
id arxiv_https___arxiv_org_abs_2310_12248
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle A PAC Learning Algorithm for LTL and Omega-regular Objectives in MDPs
Perez, Mateo
Somenzi, Fabio
Trivedi, Ashutosh
Machine Learning
Logic in Computer Science
Linear temporal logic (LTL) and omega-regular objectives -- a superset of LTL -- have seen recent use as a way to express non-Markovian objectives in reinforcement learning. We introduce a model-based probably approximately correct (PAC) learning algorithm for omega-regular objectives in Markov decision processes (MDPs). As part of the development of our algorithm, we introduce the epsilon-recurrence time: a measure of the speed at which a policy converges to the satisfaction of the omega-regular objective in the limit. We prove that our algorithm only requires a polynomial number of samples in the relevant parameters, and perform experiments which confirm our theory.
title A PAC Learning Algorithm for LTL and Omega-regular Objectives in MDPs
topic Machine Learning
Logic in Computer Science
url https://arxiv.org/abs/2310.12248