Hannover 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Hannover musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
Q: Fachverband Quantenoptik und Photonik
Q 4: Quantum gases (Fermions) I
Q 4.8: Vortrag
Montag, 9. März 2020, 12:45–13:00, e214
Experimental Quantum State Reconstruction of Few-Fermion Systems via Neural Networks — •Laurin Fischer1, Marcel Neugebauer1, Martin Gärttner2, Philipp Preiss1, Matthias Weidemüller1, and Selim Jochim1 — 1Physikalisches Institut, Universität Heidelberg, Im Neuenheimer Feld 226, 69120 Heidelberg — 2Kirchhoff-Institut für Physik, Universität Heidelberg, Im Neuenheimer Feld 227, 69120 Heidelberg
For reconstructing the state of a many-body quantum system, the number of required measurements typically scales exponentially with the system size. Machine learning techniques based on generative models have emerged in recent years as an intriguing tool to tackle this challenge. To this end, an artificial neural network is trained to encode the probability distribution of an informationally complete measurement via unsupervised learning.
This approach has been established through numerical simulations of spin models, showing promising scaling of computational resources. In order to investigate the practical feasibility of these methods, we aim to apply them to realistic experimental settings.
In this talk, we demonstrate a successful neural-network state representation of a system of few fermionic 6Li atoms in a double-well potential of optical tweezers. The training data is generated by measurements of in-situ populations in real space and correlation measurements in momentum-space, allowing us to benchmark this approach against more conventional techniques of state reconstruction, such as maximum likelihood.