Quantum optimal control of superconducting qubits based on machine-learning characterization

Implementing fast and high-fidelity quantum operations using quantum optimal control relies on having an accurate model of the quantum dynamics. Any deviations between this model and the complete dynamics of the device, such as the presence of spurious modes or pulse distortions, can degrade the performance of optimal controls in practice. Reinforcement learning eliminates the need for such an accurate quantum model by relying on fast online interactions between a controller and the quantum device. However, these model-free approaches require large data sets and yield no other useful information beyond the specific control task it was designed for. Here, we propose an experimentally simple approach to realize optimal quantum controls tailored to the device parameters and environment while explicitly characterizing this quantum system. Specifically, we use physics-inspired machine learning to infer an accurate model of the dynamics from experimental data and then optimize our experimental controls on this trained model. We demonstrate the power and feasibility of this approach by optimizing arbitrary single-qubit operations performed in parallel on superconducting transmon qubits, using both detailed numerical simulations and an experimental realization.