Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

Abstract

Recent years have seen the emergence of graph-based Knowl-
edge Bases build upon Semantic Web technologies, known as Knowledge
Graphs (KG). Popular examples are DBpedia and GeoNames. The formal system underlying these KGs provides inherent support for deductive reasoning. Growing popularity has exposed several limitations of this
ability, amongst which are scalability and uncertainty issues, as well as
coping with heterogeneous, noisy, and inconsistent data. By supplementing this form of reasoning with Machine Learning algorithms, these hurdles are much more easily overcome. Of the existing research in this area,
only a handful have been considering a Deep Neural Network. Moreover,
only one of these studies has addressed the problem of hyper-parameter
optimization, albeit under specific conditions. To contribute to this area
of research, we propose a research design that will investigate a Deep
Neural Network with optimized hyper-parameters for its effectiveness to
perform link prediction on real-world KGs.
LanguageEnglish
Title of host publicationProceedings of the ECMLPKDD 2015 Doctoral Consortium
EditorsJ Hollmén, P Papapetrou
Place of PublicationHelsinki
PublisherAalto University
Pages226-235
ISBN (Print)9789526064437
Publication statusPublished - 2015
EventEuropean Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases - Helsinki
Duration: 7 Sep 201511 Sep 2015

Conference

ConferenceEuropean Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases
Period7/09/1511/09/15

Fingerprint

Semantic Web
Learning algorithms
Learning systems
Scalability
Deep neural networks
Uncertainty

Cite this

Wilcke, W. X. (2015). Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs. In J. Hollmén, & P. Papapetrou (Eds.), Proceedings of the ECMLPKDD 2015 Doctoral Consortium (pp. 226-235). Helsinki: Aalto University.
Wilcke, W.X. / Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs. Proceedings of the ECMLPKDD 2015 Doctoral Consortium. editor / J Hollmén ; P Papapetrou. Helsinki : Aalto University, 2015. pp. 226-235
@inproceedings{09a0037128b84173b6d005ea3b3b771c,
title = "Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs",
abstract = "Recent years have seen the emergence of graph-based Knowl-edge Bases build upon Semantic Web technologies, known as KnowledgeGraphs (KG). Popular examples are DBpedia and GeoNames. The formal system underlying these KGs provides inherent support for deductive reasoning. Growing popularity has exposed several limitations of thisability, amongst which are scalability and uncertainty issues, as well ascoping with heterogeneous, noisy, and inconsistent data. By supplementing this form of reasoning with Machine Learning algorithms, these hurdles are much more easily overcome. Of the existing research in this area,only a handful have been considering a Deep Neural Network. Moreover,only one of these studies has addressed the problem of hyper-parameteroptimization, albeit under specific conditions. To contribute to this areaof research, we propose a research design that will investigate a DeepNeural Network with optimized hyper-parameters for its effectiveness toperform link prediction on real-world KGs.",
author = "W.X. Wilcke",
year = "2015",
language = "English",
isbn = "9789526064437",
pages = "226--235",
editor = "J Hollm{\'e}n and P Papapetrou",
booktitle = "Proceedings of the ECMLPKDD 2015 Doctoral Consortium",
publisher = "Aalto University",

}

Wilcke, WX 2015, Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs. in J Hollmén & P Papapetrou (eds), Proceedings of the ECMLPKDD 2015 Doctoral Consortium. Aalto University, Helsinki, pp. 226-235, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 7/09/15.

Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs. / Wilcke, W.X.

Proceedings of the ECMLPKDD 2015 Doctoral Consortium. ed. / J Hollmén; P Papapetrou. Helsinki : Aalto University, 2015. p. 226-235.

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs

AU - Wilcke, W.X.

PY - 2015

Y1 - 2015

N2 - Recent years have seen the emergence of graph-based Knowl-edge Bases build upon Semantic Web technologies, known as KnowledgeGraphs (KG). Popular examples are DBpedia and GeoNames. The formal system underlying these KGs provides inherent support for deductive reasoning. Growing popularity has exposed several limitations of thisability, amongst which are scalability and uncertainty issues, as well ascoping with heterogeneous, noisy, and inconsistent data. By supplementing this form of reasoning with Machine Learning algorithms, these hurdles are much more easily overcome. Of the existing research in this area,only a handful have been considering a Deep Neural Network. Moreover,only one of these studies has addressed the problem of hyper-parameteroptimization, albeit under specific conditions. To contribute to this areaof research, we propose a research design that will investigate a DeepNeural Network with optimized hyper-parameters for its effectiveness toperform link prediction on real-world KGs.

AB - Recent years have seen the emergence of graph-based Knowl-edge Bases build upon Semantic Web technologies, known as KnowledgeGraphs (KG). Popular examples are DBpedia and GeoNames. The formal system underlying these KGs provides inherent support for deductive reasoning. Growing popularity has exposed several limitations of thisability, amongst which are scalability and uncertainty issues, as well ascoping with heterogeneous, noisy, and inconsistent data. By supplementing this form of reasoning with Machine Learning algorithms, these hurdles are much more easily overcome. Of the existing research in this area,only a handful have been considering a Deep Neural Network. Moreover,only one of these studies has addressed the problem of hyper-parameteroptimization, albeit under specific conditions. To contribute to this areaof research, we propose a research design that will investigate a DeepNeural Network with optimized hyper-parameters for its effectiveness toperform link prediction on real-world KGs.

M3 - Conference contribution

SN - 9789526064437

SP - 226

EP - 235

BT - Proceedings of the ECMLPKDD 2015 Doctoral Consortium

A2 - Hollmén, J

A2 - Papapetrou, P

PB - Aalto University

CY - Helsinki

ER -

Wilcke WX. Learning an Optimized Deep Neural Network for Link Prediction on Knowledge Graphs. In Hollmén J, Papapetrou P, editors, Proceedings of the ECMLPKDD 2015 Doctoral Consortium. Helsinki: Aalto University. 2015. p. 226-235