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

MA: Fachverband Magnetismus

MA 7: Poster Magnetism I

MA 7.4: Poster

Montag, 9. März 2026, 09:30–12:30, P2

Optimization of Neural Quantum States for Frustrated and Chiral Spin Hamiltonians — Andreas Haller, •Stefan Liscak, Vladislav Kuchkin, Andreas Michels, and Thomas Schmidt — University of Luxembourg

Neural quantum states (NQS) offer a versatile variational Ansatz for the study of strongly correlated quantum systems, especially when dimensionality or frustration restricts the use of traditional tensor-network techniques. In order to facilitate effective ground-state searches for such Hamiltonians, we introduce a custom optimization code base. We evaluate the performance of our framework's implementations of the stochastic reconfiguration (SR) and the minimal-step SR (minSR) scheme against well-known NQS optimizers like those available in NetKet. For standard test cases, our results achieve accuracy similar to matrix product state methods (DMRG), demonstrating the competitiveness and robustness of the optimization protocols. Based on these validations, we have investigated an extension of the variational Ansatz that incorporates parameter-dependent local rotations of the computational basis directly into the optimization process. This concept is inspired by earlier findings that, following appropriate local spin rotations, skyrmionic matrix-product states show significant overlap with the simple product states. We studied the effect of adaptive basis rotations in chiral and frustrated systems on sampling requirements and convergence.

Keywords: Neural quantum states; Machine learning; Chiral magnets; Frustrated magnetism; Quantum magnetism