Bilinear dynamic mode decomposition for quantum control

Abstract Data-driven methods for establishing quantum optimal control (QOC) using time-dependent pulses tailored to specific dynamical systems and desired objectives are critical many emerging technologies. We develop a data-driven regression procedure, bilinear dynamic mode decomposition (biDMD), that leverages time-series measurements establish system identification QOC. The biDMD optimization framework is physics-informed makes use of the known underlying Hamiltonian structure. Further, can be modified model both fast slow sampling signals, latter by way stroboscopic strategies. method provides flexible, interpretable, adaptive real-time, online implementation in systems. has strong theoretical connections Koopman theory, which approximates nonlinear dynamics with linear operators. In comparison machine learning paradigms minimal data needed construct model, easily updated as new collected. demonstrate efficacy performance approach on number representative systems, showing it also matches experimental results.