Evolutionary algorithms for adaptive quantum metrology

Quantum metrology (e.g., precise displacement measurements using interferometers or precise temporal measurements using atomic clocks) aims to beat the standard quantum limit imposed by a combination of the uncertainty principle and a lack of entanglement resources. Furthermore quantum metrology aims for the ultimate quantum limit of minimal uncertainty in the variable being measured. Adaptive quantum metrology is a modified procedure that uses feedback and control to reach for the same limit while avoiding the overhead of having to perform (effectively entangling) joint measurements on the output (photons in interferometry and atoms in clocks). Adaptive procedures are hard to devise even in ideal cases such as noiseless isolated metrological systems. I show how we employ machine learning in the form of evolutionary algorithms to devise adaptive procedures that are superior to those devised by human minds. Our procedure for devising adaptive feedback procedures is effective even for noisy cases. This work is important not only for quantum metrology but also as a first step to new, powerful methods for quantum control.