Sitzungen | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

FM: Fall Meeting

FM 42: Poster: Quantum Computation

FM 42.7: Poster

Dienstag, 24. September 2019, 16:30–18:30, Tents

Neural Networks for Reconstructing Quantum Gas Microscope Images — •Bastian Lunow, Niklas Käming, Andreas Kerkmann, Michael Hagemann, Mathis Fischer, Klaus Sengstock, and Christof Weitenberg — ILP Institut für Laserphysik, Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany

Quantum gas microscopes allow to image single atoms in optical lattices and therefore give a new window into many-body physics with microscopic access. One challenge is to reconstruct the occupation of the atoms on the lattice from finite fluorescence signal, even when the lattice sites are not fully optically resolved by the imaging system. Reconstruction algorithms allow to reconstruct them anyway. It has already been shown that this can be done with conventional methods, but we hope to achieve better results by using neural nets which are known to be efficient in extracting features from image data. Here we investigate the use of machine learning techniques in order to improve upon the conventional reconstruction algorithms with the promise to yield a faithful and faster reconstruction already for lower fluorescence counts. Using the concrete example of a triangular pinning lattice, we focus on realistic parameters for lithium atoms. Fruitful architectures include variational autoencoders and the use of weight initializations. These approaches would open a path to applying quantum gas microscopes to more general setup and more exotic atomic species.