In this paper, published in Physical Review A, we show how to greatly improve success at solving Constraint Satisfaction Problems on a quantum computer by using a learned schedule, instead of the standard linear ramps. The technique actually improves as the problem gets larger and more difficult, allowing classical machines to learn optimizations that can aid quantum machines.
Read the published version.