Umeå University's logo

umu.sePublikasjoner
Endre søk
RefereraExporteraLink to record
Permanent link

Direct link
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annet format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annet språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf
Similarity-based link prediction from modular compression of network flows
Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för fysik.ORCID-id: 0000-0001-7881-2496
Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för fysik.ORCID-id: 0000-0003-0124-1909
Center for Artificial Intelligence and Data Science, University of Würzburg, Germany.
Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för fysik.ORCID-id: 0000-0002-7181-9940
2022 (engelsk)Inngår i: Proceedings of the First Learning on Graphs Conference, ML Research Press , 2022, s. 52:1-52:18Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

Node similarity scores are a foundation for machine learning in graphs for clustering, node classification, anomaly detection, and link prediction with applications in biological systems, information networks, and recommender systems. Recent works on link prediction use vector space embeddings to calculate node similarities in undirected networks with good performance. Still, they have several disadvantages: limited interpretability, need for hyperparameter tuning, manual model fitting through dimensionality reduction, and poor performance from symmetric similarities in directed link prediction. We propose MapSim, an information-theoretic measure to assess node similarities based on modular compression of network flows. Unlike vector space embeddings, MapSim represents nodes in a discrete, non-metric space of communities and yields asymmetric similarities in an unsupervised fashion. We compare MapSim on a link prediction task to popular embedding-based algorithms across 47 networks and find that MapSim's average performance across all networks is more than 7% higher than its closest competitor, outperforming all embedding methods in 11 of the 47 networks. Our method demonstrates the potential of compression-based approaches in graph representation learning, with promising applications in other graph learning tasks.

sted, utgiver, år, opplag, sider
ML Research Press , 2022. s. 52:1-52:18
Serie
Proceedings of Machine Learning Research, E-ISSN 2640-3498 ; 198
HSV kategori
Identifikatorer
URN: urn:nbn:se:umu:diva-212276Scopus ID: 2-s2.0-85164537856OAI: oai:DiVA.org:umu-212276DiVA, id: diva2:1783301
Konferanse
LOG 2022, 1st Learning on Graphs Conference, Virtual, December9-12, 2022
Forskningsfinansiär
Wallenberg AI, Autonomous Systems and Software Program (WASP)Knut and Alice Wallenberg FoundationSwedish Research Council, 2016-00796Tilgjengelig fra: 2023-07-20 Laget: 2023-07-20 Sist oppdatert: 2023-07-20bibliografisk kontrollert

Open Access i DiVA

Fulltekst mangler i DiVA

Andre lenker

ScopusPublisher's full text

Person

Blöcker, ChristopherSmiljanic, JelenaRosvall, Martin

Søk i DiVA

Av forfatter/redaktør
Blöcker, ChristopherSmiljanic, JelenaRosvall, Martin
Av organisasjonen

Søk utenfor DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric

urn-nbn
Totalt: 154 treff
RefereraExporteraLink to record
Permanent link

Direct link
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annet format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annet språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf