Curriculum reinforcement learning via constrained optimal transport

dc.contributorAalto-yliopistofi
dc.contributorAalto Universityen
dc.contributor.authorKlink, Pascalen_US
dc.contributor.authorYang, Haoyien_US
dc.contributor.authorD'Eramo, Carloen_US
dc.contributor.authorPajarinen, Jonien_US
dc.contributor.authorPeters, Janen_US
dc.contributor.departmentDepartment of Electrical Engineering and Automationen
dc.contributor.groupauthorRobot Learningen
dc.contributor.organizationTechnische Universität Darmstadten_US
dc.date.accessioned2023-01-18T09:21:18Z
dc.date.available2023-01-18T09:21:18Z
dc.date.issued2022en_US
dc.description.abstractCurriculum reinforcement learning (CRL) allows solving complex tasks by generating a tailored sequence of learning tasks, starting from easy ones and subsequently increasing their difficulty. Although the potential of curricula in RL has been clearly shown in a variety of works, it is less clear how to generate them for a given learning environment, resulting in a variety of methods aiming to automate this task. In this work, we focus on the idea of framing curricula as interpolations between task distributions, which has previously been shown to be a viable approach to CRL. Identifying key issues of existing methods, we frame the generation of a curriculum as a constrained optimal transport problem between task distributions. Benchmarks show that this way of curriculum generation can improve upon existing CRL methods, yielding high performance in a variety of tasks with different characteristics.en
dc.description.versionPeer revieweden
dc.format.extent18
dc.format.mimetypeapplication/pdfen_US
dc.identifier.citationKlink, P, Yang, H, D'Eramo, C, Pajarinen, J & Peters, J 2022, Curriculum reinforcement learning via constrained optimal transport . in Proceedings of the 39th International Conference on Machine Learning . Proceedings of Machine Learning Research, vol. 162, JMLR, International Conference on Machine Learning, Baltimore, Maryland, United States, 17/07/2022 . < https://proceedings.mlr.press/v162/klink22a/klink22a.pdf >en
dc.identifier.issn2640-3498
dc.identifier.otherPURE UUID: 3b95c60f-85ac-4dab-aacc-4524650b5772en_US
dc.identifier.otherPURE ITEMURL: https://research.aalto.fi/en/publications/3b95c60f-85ac-4dab-aacc-4524650b5772en_US
dc.identifier.otherPURE LINK: https://proceedings.mlr.press/v162/klink22a/klink22a.pdfen_US
dc.identifier.otherPURE FILEURL: https://research.aalto.fi/files/96482753/klink22a.pdfen_US
dc.identifier.urihttps://aaltodoc.aalto.fi/handle/123456789/118829
dc.identifier.urnURN:NBN:fi:aalto-202301181185
dc.language.isoenen
dc.publisherPMLR
dc.relation.ispartofInternational Conference on Machine Learningen
dc.relation.ispartofseriesProceedings of the 39th International Conference on Machine Learningen
dc.relation.ispartofseriesProceedings of Machine Learning Researchen
dc.relation.ispartofseriesVolume 162en
dc.rightsopenAccessen
dc.titleCurriculum reinforcement learning via constrained optimal transporten
dc.typeConference article in proceedingsfi
dc.type.versionpublishedVersion
Files