Show simple item record

dc.contributor.authorKampffmeyer, Michael C.
dc.contributor.authorChen, Yinbo
dc.contributor.authorLiang, Xiaodan
dc.contributor.authorWang, Hao
dc.contributor.authorZhang, Yujia
dc.contributor.authorXing, Eric P.
dc.date.accessioned2020-03-04T11:54:01Z
dc.date.available2020-03-04T11:54:01Z
dc.date.issued2019
dc.description.abstractGraph convolutional neural networks have recently shown great potential for the task of zero-shot learning. These models are highly sample efficient as related concepts in the graph structure share statistical strength allowing generalization to new classes when faced with a lack of data. However, multi-layer architectures, which are required to propagate knowledge to distant nodes in the graph, dilute the knowledge by performing extensive Laplacian smoothing at each layer and thereby consequently decrease performance. In order to still enjoy the benefit brought by the graph structure while preventing dilution of knowledge from distant nodes, we propose a Dense Graph Propagation (DGP) module with carefully designed direct links among distant nodes. DGP allows us to exploit the hierarchical graph structure of the knowledge graph through additional connections. These connections are added based on a node's relationship to its ancestors and descendants. A weighting scheme is further used to weigh their contribution depending on the distance to the node to improve information propagation in the graph. Combined with finetuning of the representations in a two-stage training approach our method outperforms state-of-the-art zero-shot learning approaches.en_US
dc.description© 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.en_US
dc.identifier.citationKampffmeyer MC, Chen, Liang X, Wang H, Zhang Y, Xing EP. Rethinking knowledge graph propagation for zero-shot learning. Computer Vision and Pattern Recognition. 2019en_US
dc.identifier.cristinIDFRIDAID 1749901
dc.identifier.doihttps://doi.org/10.1109/CVPR.2019.01175
dc.identifier.issn1063-6919
dc.identifier.urihttps://hdl.handle.net/10037/17600
dc.language.isoengen_US
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)en_US
dc.relation.journalComputer Vision and Pattern Recognition
dc.relation.projectIDNorges forskningsråd: 239844en_US
dc.relation.projectIDinfo:eu-repo/grantAgreement/RCN/?/239844/Norway/?/?/en_US
dc.rights.accessRightsopenAccessen_US
dc.rights.holder© Copyright 2019 IEEE – All rights reserved.en_US
dc.subjectVDP::Mathematics and natural science: 400en_US
dc.subjectVDP::Matematikk og Naturvitenskap: 400en_US
dc.titleRethinking knowledge graph propagation for zero-shot learningen_US
dc.type.versionacceptedVersionen_US
dc.typeJournal articleen_US
dc.typeTidsskriftartikkelen_US
dc.typePeer revieweden_US


File(s) in this item

Thumbnail

This item appears in the following collection(s)

Show simple item record