Automated Design Simplification of Quantum Program Synthesis

Quantum computers are a new and emerging technology that offer promises of being able to outperform classical machines. However, they differ from classical machines so much that they provide unique challenges to development. Working on quantum machines is currently very difficult, requiring a large amount of expertise in a great deal of areas. In order to facilitate practical software engineering methods it will be necessary to greatly simplify this process. To provide this process simplification we identify automation methods and approaches that can perform steps of quantum program compilation to greatly reduce the need for human expertise.

The first contribution looks at integrating an existing classical method into the quantum model. This is done through the application of a Genetic Improvement algorithm. The second contribution looks at modelling the quantum compilation problem in a way compatible with a classical model. This is done through the generation of a Planning Domain Definition Language (PDDL) model. The third and final contribution looks at simplifying the building of a compilation stack. This is done by using a neural network to make decisions about what steps to add to the compilation stack.

The results of this show a set of automated methods that produce error rates competitive with the standard quantum compilation methods. In addition, these methods require much less expertise about specific quantum hardware or the quantum compilation stack and are built to be compatible with the current IBM Quantum Experience software stack.