Experimental graybox quantum identification and control

Modeling and controlling a quantum system are crucial for developing useful quantum devices to overcome errors that come from environmental noise or fabrication imperfections. Many methods have been developed to tackle this challenge. However, they are either limited by the system model such as traditional curve fitting, or lack of physics insights such as machine learning methods. Here we experimentally demonstrate a 'graybox' method on a reconfigurable photonic circuit. Our results show that the 'graybox' method outperforms the traditional model fitting method while holding the capability of providing physics insights. This new method is effective in modeling devices whose properties cannot be measured directly and can be applied to time-dependent and open quantum systems.