Machine learning for adaptive quantum measurement

Machine learning is an automated process for creating and enhancing algorithms and has had revolutionary impact in data mining and pattern recognition. Quantum information presents a new challenge because the input is a quantum state rather than bit strings, but machine learning could nonetheless be used to devise and improve quantum information tasks and protocols. Here we develop an automated method, based on machine learning, to generate adaptive feedback measurement protocols. We apply our technique to adaptive quantum phase measurement, which is important for applications such as ultra-precise atomic clocks and gravitational wave detection. Our protocols, autonomously constructed by machine learning, outperform the best known adaptive measurement scheme for estimating an unknown interferometric phase shift.