Learning sparse and meaningful representations through embodiment
DC Element | Wert | Sprache |
---|---|---|
dc.contributor.author | Clay, Viviane | |
dc.contributor.author | Koenig, Peter | |
dc.contributor.author | Kuehnberger, Kai-Uwe | |
dc.contributor.author | Pipa, Gordon | |
dc.date.accessioned | 2021-12-23T16:13:08Z | - |
dc.date.available | 2021-12-23T16:13:08Z | - |
dc.date.issued | 2021 | |
dc.identifier.issn | 08936080 | |
dc.identifier.uri | https://osnascholar.ub.uni-osnabrueck.de/handle/unios/10423 | - |
dc.description.abstract | How do humans acquire a meaningful understanding of the world with little to no supervision or semantic labels provided by the environment? Here we investigate embodiment with a closed loop between action and perception as one key component in this process. We take a close look at the representations learned by a deep reinforcement learning agent that is trained with high-dimensional visual observations collected in a 3D environment with very sparse rewards. We show that this agent learns stable representations of meaningful concepts such as doors without receiving any semantic labels. Our results show that the agent learns to represent the action relevant information, extracted from a simulated camera stream, in a wide variety of sparse activation patterns. The quality of the representations learned shows the strength of embodied learning and its advantages over fully supervised approaches. (C) 2020 The Authors. Published by Elsevier Ltd. | |
dc.description.sponsorship | Deutsche Forschungs Gesellschaft (DFG), GermanyGerman Research Foundation (DFG) [GRK2340]; The project was financed by the funds of a research training group provided by the Deutsche Forschungs Gesellschaft (DFG), Germany (GRK2340). | |
dc.language.iso | en | |
dc.publisher | PERGAMON-ELSEVIER SCIENCE LTD | |
dc.relation.ispartof | NEURAL NETWORKS | |
dc.subject | Computer Science | |
dc.subject | Computer Science, Artificial Intelligence | |
dc.subject | Deep learning | |
dc.subject | Embodied cognition | |
dc.subject | Embodiment | |
dc.subject | Neurosciences | |
dc.subject | Neurosciences & Neurology | |
dc.subject | PERCEPTION | |
dc.subject | Reinforcement learning | |
dc.subject | Representation learning | |
dc.subject | SEE | |
dc.subject | Sparse coding | |
dc.title | Learning sparse and meaningful representations through embodiment | |
dc.type | journal article | |
dc.identifier.doi | 10.1016/j.neunet.2020.11.004 | |
dc.identifier.isi | ISI:000603296800003 | |
dc.description.volume | 134 | |
dc.description.startpage | 23 | |
dc.description.endpage | 41 | |
dc.contributor.orcid | 0000-0001-9152-0666 | |
dc.contributor.orcid | 0000-0003-3654-5267 | |
dc.contributor.orcid | 0000-0003-1626-0598 | |
dc.contributor.researcherid | AAR-8308-2021 | |
dc.identifier.eissn | 18792782 | |
dc.publisher.place | THE BOULEVARD, LANGFORD LANE, KIDLINGTON, OXFORD OX5 1GB, ENGLAND | |
dcterms.isPartOf.abbreviation | Neural Netw. | |
dcterms.oaStatus | hybrid | |
crisitem.author.dept | Institut für Kognitionswissenschaft | - |
crisitem.author.dept | FB 05 - Biologie/Chemie | - |
crisitem.author.dept | Institut für Kognitionswissenschaft | - |
crisitem.author.dept | Institut für Kognitionswissenschaft | - |
crisitem.author.deptid | institute28 | - |
crisitem.author.deptid | fb05 | - |
crisitem.author.deptid | institute28 | - |
crisitem.author.deptid | institute28 | - |
crisitem.author.orcid | 0000-0003-3654-5267 | - |
crisitem.author.orcid | 0000-0003-1626-0598 | - |
crisitem.author.orcid | 0000-0002-3416-2652 | - |
crisitem.author.parentorg | FB 08 - Humanwissenschaften | - |
crisitem.author.parentorg | Universität Osnabrück | - |
crisitem.author.parentorg | FB 08 - Humanwissenschaften | - |
crisitem.author.parentorg | FB 08 - Humanwissenschaften | - |
crisitem.author.grandparentorg | Universität Osnabrück | - |
crisitem.author.grandparentorg | Universität Osnabrück | - |
crisitem.author.grandparentorg | Universität Osnabrück | - |
crisitem.author.netid | KoPe298 | - |
crisitem.author.netid | KuKa032 | - |
crisitem.author.netid | PiGo340 | - |
Seitenaufrufe
9
Letzte Woche
0
0
Letzter Monat
0
0
geprüft am 14.05.2024