Bonn 2025 – wissenschaftliches Programm

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

QI: Fachverband Quanteninformation

QI 34: Quantum Control I

QI 34.4: Vortrag

Donnerstag, 13. März 2025, 15:15–15:30, HS II

Neural-network-based preparation of quantum state families: Theory and experiment — Hector Hutin¹, •Pavlo Bilous², Florian Marquardt^{2, 3}, and Benjamin Huard¹ — ¹Ecole Normale Supérieure de Lyon, CNRS, Laboratoire de Physique, 69342 Lyon, France — ²Max Planck Institute for the Science of Light, Staudtstr. 2, 91058 Erlangen, Germany — ³Department of Physics, Friedrich-Alexander-Universität Erlangen-Nürnberg, 91058 Erlangen, Germany

Fast preparation of quantum states is a crucial ingredient for scaling up quantum computing devices. Along with the established techniques like GRadient Ascent Pulse Engineering (GRAPE), the neural-network (NN) methods are being increasingly employed for this task. However, using a NN for preparation of a fixed single quantum state implies a very slow training from scratch once a different quantum state is required.

We present a way to teach a NN quantum state preparation for a continuous family of states instead of a single state. Once trained on a random selection from the family, the NN is able to predict control signals for any quantum state from the family. Building up on the original theoretical proposal from Ref. [1], we introduced further theoretical developments and demonstrated the method experimentally for Schrödinger cat states [2]. The method can be useful e.g. for implementation of parametrized quantum gates requiring fast switching between quantum states.

[1] F. Sauvage and F. Mintert, Phys. Rev. Lett. 129, 050507 (2022).

[2] H. Hutin, P. Bilous et. al. arxiv.org:2409.05557 (2024).

Keywords: Quantum state preparation; Quantum control; Neural networks; Machine learning; Schrödinger cat states