Parts | Days | Selection | Search | Updates | Downloads | Help

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 4: AKPIK III: Simulation & Application

AKPIK 4.7: Talk

Thursday, March 18, 2021, 17:30–17:45, AKPIKa

Classification of spin qubit detection events with neural networks — •Tom Struck¹, Javed Lindner¹, Arne Hollmann¹, Lars R. Schreiber¹, Floyd Schauer², Andreas Schmidbauer², and Dominique Bougeard² — ¹JARA-Fit Institute for Quantum Information, Forschungszentrum Jülich GmbH and RWTH Aachen University, Aachen, Germany — ²Institut für Experimentelle und Angewandte Physik, Universität Regensburg, Regensburg Germany

Fast and accurate detection of a qubit state is essential for quantum information processing, in particular for quantum error correction. Here, we detect the state of a single electron spin, localized in a Si/SiGe quantum dot, in a single shot measurement using a single-electron transistor [1]. We investigate the capability of a neural network to classify the experimental signal traces into spin-up and -down events [2] and compare the network performance to a state-of-the-art Bayesian inference filter, which is theoretically optimal for signals with Gaussian noise. We find that the neural network can outperform the Bayesian filter on experimentally recorded data. The network can be made robust to setup-variations by training with proper synthetic traces.

[1] T. Struck et al., npj Quantum Inf. 6, 40 (2020).

[2] T. Struck et al., arXiv:2012.04686 (2020).