Supervised Learning Enhanced Quantum Circuit Transformation

A quantum circuit transformation (QCT) is required when executing a program in real processing unit (QPU). Through inserting auxiliary SWAP gates, QCT algorithm transforms to one that satisfies the connectivity constraint imposed by QPU. Due non-negligible gate error and limited qubit coherence time of QPU, algorithms which minimize number or depth maximize fidelity output circuits are urgent need. Unfortunately, finding optimized transformations often involves exhaustive searches, extremely time-consuming not practical for most circuits. In this paper, we propose framework uses policy artificial neural network (ANN) trained supervised learning on shallow help existing select promising gate. ANNs can be off-line distributed way. The ANN easily incorporated into without bringing too much overhead complexity. Exemplary embeddings target demonstrate performance consistently improved QPUs with various structures random realistic