Quantum compiling by deep reinforcement learning

The architecture of circuital quantum computers requires computing layers devoted to compiling high-level algorithms into lower-level circuits gates. general problem is approximate any unitary transformation that describes the computation, as a sequence elements selected from finite base universal existence an approximating one qubit gates guaranteed by Solovay-Kitaev theorem, which implies sub-optimal establish it explicitly. Since may require significantly different gate sequences, depending on considered, such great complexity and does not admit efficient algorithm. Therefore, traditional approaches are time-consuming tasks, unsuitable be employed during computation. We exploit deep reinforcement learning method alternative strategy, has trade-off between search time exploitation time. Deep allows creating single-qubit operations in real time, after arbitrary long training period strategy for sequences operators built. based fast computation times, could principle exploited real-time compiling.