Our recently published work shows the first experimental demonstration of a new quantum system identification and control technique that adds the benefit of machine learning to the physical insight of classical fitting models, namely “graybox quantum control on a quantum device”. The method overcomes the main limitation of machine learning, which is not providing information on the physics of a quantum process, while outperforming traditional model fitting approaches. This technique adds to the flexibility and power of machine learning by enabling the preparation of unitary gates and Hamiltonians, which is not possible with standard machine learning methods, such as neural networks. Applied to our designed and fabricated quantum photonic device, the results show superior performance in terms of the experimentally measured controlled performance. Moreover, the method uncovered a new enhanced physical model that goes beyond the standard well-known device model.
https://www.nature.com/articles/s41534-023-00795-5