Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
QI: Fachverband Quanteninformation
QI 16: Quantum Software
QI 16.5: Vortrag
Donnerstag, 12. März 2026, 10:45–11:00, BEY/0245
Real-time adaptive quantum error correction by model-free multi-agent learning — •Manuel Guatto1,2, Francesco Preti1, Michael Schilling1,2, Tommaso Calarco1,2,3, Francisco Cardenas-Lopez1, and Felix Motzoi1,2 — 1Forschungszentrum Jülich GmbH, Peter Grünberg Institute, Quantum Control (PGI-8), 52425 Jülich, Germany — 2Institute for Theoretical Physics, University of Cologne, D-50937 Cologne, Germany — 3Dipartimento di Fisica e Astronomia, Università di Bologna, 40127 Bologna, Italy
Can Quantum Error Correction (QEC) adapt in real time to changing noise? We show that it can. We introduce a two-level reinforcement-learning framework that learns QEC from scratch and adapts it on the fly. At the first level, a model-free multi-agent RL system automatically discovers full QEC cycles*encoding, stabilizer measurements, and recovery*using only orthogonality constraints and no prior knowledge of the device. Using the stabilizer formalism, we demonstrate that it can generate new QEC codes tailored to multi-level quantum architectures. At the second level, we present BRAVE (Bandit Retraining for Adaptive Variational Error correction), an efficient algorithm that continuously retunes the variational layer to track time-dependent noise with minimal retraining. Combined, these methods yield more than an order-of-magnitude improvement in logical fidelity under time-varying bit- and phase-flip noise compared to standard QEC schemes.
Keywords: Quantum Error correction; Reinforcement Learning; Noise models