Szczegóły publikacji
Opis bibliograficzny
Estimating data split error for training knowledge graph embeddings / Aleksander SUCHORAB, Igor WOJNICKI, Andrzej BIELECKI // W: Computational Science – ICCS 2026 : 26th International Conference : Hamburg, Germany, June 29–July 1, 2026 : proceedings , Pt. 2 / eds. Philipp Neumann [et al.]. — Cham : Springer, cop. 2026. — ( Lecture Notes in Computer Science ; ISSN 0302-9743 ; LNCS 16784 ). — ISBN: 978-3-032-29923-9; e-ISBN: 978-3-032-29924-6. — S. 506–514. — Bibliogr., Abstr. — Publikacja dostępna online od: 2026-06-27
Autorzy (3)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 168886 |
|---|---|
| Data dodania do BaDAP | 2026-08-28 |
| DOI | 10.1007/978-3-032-29924-6_46 |
| Rok publikacji | 2026 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Springer |
| Konferencja | International Conference on Computational Science 2026 |
| Czasopismo/seria | Lecture Notes in Computer Science |
Abstract
When graph data is divided into a train-test split, which is common in machine learning, it alters the graph structure. In particular, some splits may result in a training set that is unable to distinguish certain groups of vertices as a consequence of graph topology. It can lead to an increased number of errors in evaluation when the test set requires a distinction between those groups but the training set provides none. We propose methods to identify, explain, and estimate the split error of a graph dataset with a particular assignment of triples into splits. The method uses graph entropy differences, based on automorphism of labeled graphs. We examine the approach on a variety of knowledge graph embedding models and datasets. Various datasets were found to have differing potential for split errors, depending on their structure.
