Full metadata record

DC Field Value Language
dc.contributor.authorSong, Wonil-
dc.contributor.authorJeon, Sangryul-
dc.contributor.authorChoi, Hyesong-
dc.contributor.authorSohn, Kwanghoon-
dc.contributor.authorMin, Dongbo-
dc.date.accessioned2024-01-19T08:30:54Z-
dc.date.available2024-01-19T08:30:54Z-
dc.date.created2023-09-07-
dc.date.issued2023-11-
dc.identifier.issn0957-4174-
dc.identifier.urihttps://pubs.kist.re.kr/handle/201004/113150-
dc.description.abstractTypically, hierarchical reinforcement learning (RL) requires skills that are applicable to various downstream tasks. Although several recent studies have proposed the supervised and unsupervised learning of such skills, the learned skills are often entangled, which hinders their interpretation. To alleviate this, we propose a novel method to use weak labels for learning disentangled skills from the continuous latent representations of trajectories. To this end, we extended a trajectory variational autoencoder (VAE) to impose an inductive bias using weak labels, which explicitly enforces the disentangling of the trajectory representations into factors of interest intended for the model to learn. Using the latent representations as skills, a skill-based policy network is trained to generate trajectories similar to the learned decoder of the trajectory VAE. Furthermore, using the disentangled skill, we propose a skill repetition that can expand the entire trajectories generated by the policy at test time, resulting in an effective planning strategy. Experiments were performed on several challenging navigation tasks in mazes, and the results demonstrate the effectiveness of our method at solving hierarchical RL problems even with a long horizon and sparse rewards.-
dc.languageEnglish-
dc.publisherPergamon Press Ltd.-
dc.titleLearning disentangled skills for hierarchical reinforcement learning through trajectory autoencoder with weak labels-
dc.typeArticle-
dc.identifier.doi10.1016/j.eswa.2023.120625-
dc.description.journalClass1-
dc.identifier.bibliographicCitationExpert Systems with Applications, v.230-
dc.citation.titleExpert Systems with Applications-
dc.citation.volume230-
dc.description.isOpenAccessN-
dc.description.journalRegisteredClassscie-
dc.description.journalRegisteredClassscopus-
dc.identifier.wosid001054200000001-
dc.identifier.scopusid2-s2.0-85161724253-
dc.relation.journalWebOfScienceCategoryComputer Science, Artificial Intelligence-
dc.relation.journalWebOfScienceCategoryEngineering, Electrical & Electronic-
dc.relation.journalWebOfScienceCategoryOperations Research & Management Science-
dc.relation.journalResearchAreaComputer Science-
dc.relation.journalResearchAreaEngineering-
dc.relation.journalResearchAreaOperations Research & Management Science-
dc.type.docTypeArticle-
dc.subject.keywordPlusEMERGENCY-DEPARTMENT VISITS-
dc.subject.keywordPlusDEEP-
dc.subject.keywordAuthorDeep reinforcement learning-
dc.subject.keywordAuthorHierarchical reinforcement learning-
dc.subject.keywordAuthorSkill learning-
dc.subject.keywordAuthorVariational autoencoder-
dc.subject.keywordAuthorDisentangled representation-
dc.subject.keywordAuthorWeak label-
dc.subject.keywordAuthorPlanning-
Appears in Collections:
KIST Article > 2023
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML

qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE