In the hardware implementation of quantum computation, the control pulses for quantum states and gate operations must be calibrated to correct errors induced by unknown drifts of system parameters. The calibration can be iteratively done by checking the control precision, but the required randomized benchmarking or quantum tomography is experimentally expensive. In this paper, we propose that the control calibration and the quantum tomography can be taken as two separate but collaborative learning processes. Combining the gradient-descent pulse engineering and the adaptive tomography, the resulting c-GRAPE algorithm can substantially reduce the high cost of measurements without sacrificing the control accuracy. The effectiveness is demonstrated by numerical simulation examples.

中文翻译：

---

高精度量子控制和断层扫描的协同学习

在量子计算的硬件实现中，必须校准量子态和门操作的控制脉冲，以纠正由系统参数未知漂移引起的误差。校准可以通过检查控制精度来迭代完成，但所需的随机基准或量子断层扫描在实验上是昂贵的。在本文中，我们建议将控制校准和量子断层扫描视为两个独立但协作的学习过程。结合梯度下降脉冲工程和自适应断层扫描，由此产生的 c-GRAPE 算法可以在不牺牲控制精度的情况下大大降低测量的高成本。数值模拟实例证明了其有效性。