Parts | Days | Selection | Search | Updates | Downloads | Help
Q: Fachverband Quantenoptik und Photonik
Q 3: Quantum Computing and Simulation (joint session Q/QI)
Q 3.6: Talk
Monday, March 6, 2023, 12:15–12:30, E214
Physical computing with a superfluid — Maurus Hans1, •Elinor Kath1, Marius Sparn1, Nikolas Liebster1, Felix Draxler2, Helmut Strobel1, and Markus Oberthaler1 — 1Kirchhoff-Institut für Physik, Universität Heidelberg, Germany — 2Interdisziplinäres Zentrum für Wissenschaftliches Rechnen, Universität Heidelberg, Germany
We report on the implementation of a hybrid neural network with a physical system. As a proof-of-concept we implement the regression and interpolation of a non-linear, one-dimensional function. A digital micromirror device is used to prepare an elongated atomic cloud and encode input values by imprinting a phase profile onto the superfluid. Its non-linear response is detected by the observation of the density distribution, from which the output value is generated by a trained linear layer. We compare the performance of this hybrid neural network for different parameters and give an outlook for further directions.