EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Uncertainty quantification in graph neural networks with shallow ensembles

Tirtha Vinchurkar, Kareem Abdelmaqsoud, John R Kitchin

Machine Learning: Science and Technology · 30.12.2025 · 6 (4) · 045007

DOI: 10.1088/2632-2153/ae0bf0 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Machine Learning: Science and Technology, online erschienen am 10.10.2025. Verfasst von Tirtha Vinchurkar und 2 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref) und nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Machine-learned potentials (MLPs) have revolutionized materials discovery by providing accurate and efficient predictions of molecular and material properties. Graph neural networks (GNNs) have emerged as a state-of-the-art approach due to their ability to capture complex atomic interactions. However, GNNs often produce unreliable predictions when encountering out-of-domain data and it is difficult to identify when that happens. To address this challenge, we explore uncertainty quantification (UQ) techniques, focusing on direct propagation of shallow ensembles (DPOSEs) as a computationally efficient alternative to deep ensembles. By integrating DPOSE into the SchNet model, we assess its ability to provide reliable uncertainty estimates across several density functional theory datasets, including QM9, OC20, and Gold dataset. Our findings often demonstrate that DPOSE successfully distinguishes between in-domain and out-of-domain samples, exhibiting higher uncertainty for unobserved molecule and material classes. This work highlights the potential of lightweight UQ methods in improving the robustness of GNN-based materials modeling and lays the foundation for future integration with active learning strategies.

Zitieren

APA

Vinchurkar, T., Abdelmaqsoud, K., & Kitchin, J. R. (2025). Uncertainty quantification in graph neural networks with shallow ensembles. Machine Learning: Science and Technology, 6(4), 045007. https://doi.org/10.1088/2632-2153/ae0bf0

Vancouver

Vinchurkar T, Abdelmaqsoud K, Kitchin JR. Uncertainty quantification in graph neural networks with shallow ensembles. Machine Learning: Science and Technology. 2025;6(4):045007. doi:10.1088/2632-2153/ae0bf0

BibTeX
@article{Vinchurkar2025,
  author = {Vinchurkar, Tirtha and Abdelmaqsoud, Kareem and Kitchin, John R},
  title = {Uncertainty quantification in graph neural networks with shallow ensembles},
  journal = {Machine Learning: Science and Technology},
  year = {2025},
  volume = {6},
  number = {4},
  pages = {045007},
  doi = {10.1088/2632-2153/ae0bf0}
}
Kostenlos starten →