Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

QI: Fachverband Quanteninformation

QI 21: Quantum Information: Concepts and Methods III

QI 21.7: Vortrag

Freitag, 13. März 2026, 11:45–12:00, BEY/0245

Efficient Entanglement Quantification with a Graph Neural Network — •Susanna Bräu, Martina Jung, and Martin Gärttner — Institut für Festkörpertheorie und Optik, Friedrich-Schiller-Universität Jena, Max-Wien-Platz 1, 07743 Jena

Entanglement is a fundamental feature of quantum mechanics, yet quantifying it - using measures such as entanglement entropy - generally requires reconstruction of the full quantum state. However, this is infeasible for larger systems, limiting the accessible system sizes. In this work, we predict quantum correlation measures for many-body spin systems from measurement snapshots with a supervised machine learning approach, avoiding full quantum state tomography. Our approach uses a permutation-invariant graph neural network (GNN), which scales linearly with the system size. To improve the scaling with the number of snapshots, we implemented a mini-set architecture, that divides the input into smaller subsets processed in parallel. This modified architecture enables entanglement prediction for larger data sets, potentially allowing for more accurate predictions than the traditional architecture without increasing the number of parameters in the network significantly. Furthermore, we aim to extend the approach from quantum many-body systems to continuous variable systems.

Keywords: Machine learning; Entanglement quantification; Graph neural networks; Quantum many-body systems; Quantum simulation