Collaborative Learning of High-Precision Quantum Control and Tomography

摘要

High-precision control of quantum states and gate operations is essential to the hardware implementation of quantum computation. Recently, online calibration has become an important tool for correcting errors induced by parameter shifts or environmental noises in the underlying quantum control systems. However, the experimental cost for acquiring information through quantum tomography (for state or gate reconstruction) is very high, especially when many iterations are to be done. In this paper, we propose a novel scheme that integrates the gradient-descent optimization of quantum control pulses with the adaptive learning of quantum tomography as two interactive processes, which updates the control iteratively with the progressively refined state tomography. This scheme, which we call c-GRAPE, can greatly improve the calibration efficiency by substantial reduction the experimental cost for tomography without sacrificing the control precision.