Relational Constraints for Metric Learning on Relational Data

Jiajun Pan 1, 2 Hoel Le Capitaine 1, 2 Philippe Leray 1, 2
2 DUKe - Data User Knowledge
LS2N - Laboratoire des Sciences du Numérique de Nantes
Abstract : Most of metric learning approaches are dedicated to be applied on data described by feature vectors, with some notable exceptions such as times series, trees or graphs. The objective of this paper is to propose a metric learning algorithm that specifically considers relational data. The proposed approach can take benefit from both the topological structure of the data and supervised labels. For selecting relative constraints representing the relational information, we introduce a link-strength function that measures the strength of relationship links between entities by the side-information of their common parents. We show the performance of the proposed method with two different classical metric learning algorithms, which are ITML (Information Theoretic Metric Learning) and LSML (Least Squares Metric Learning), and test on several real-world datasets. Experimental results show that using relational information improves the quality of the learned metric.
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-02017253
Contributor : Hoel Le Capitaine <>
Submitted on : Wednesday, February 13, 2019 - 9:51:24 AM
Last modification on : Tuesday, April 9, 2019 - 4:31:30 PM

File

starai.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02017253, version 1

Collections

Citation

Jiajun Pan, Hoel Le Capitaine, Philippe Leray. Relational Constraints for Metric Learning on Relational Data. Eighth International Workshop on Statistical Relational AI, IJCAI, Jul 2018, Stockholm, Sweden. ⟨hal-02017253⟩

Share

Metrics

Record views

12

Files downloads

8