Machine learning for hard quantum control

Although quantum control is remarkably easy for many applications, leading to huge success using greedy optimization algorithms, quantum control for quantum computing changes the game by imposing strong constraints on quantum preparation and measurement such as fidelity and run-time. We have developed optimization techniques based on supervised and reinforcement learning, including differential evolution and recently MATLAB's easy-to-use GlobalSearch, for dealing with highly constrained quantum control, which we call “hard quantum control”. I review our results on adaptive quantum metrology and designing quantum gates and discuss our path forward for hard quantum control in real-world settings.