Quantum assisted quantum compiling

摘要: Compiling quantum algorithms for near-term computers (accounting connectivity and native gate alphabets) is a major challenge that has received significant attention both by industry academia. Avoiding the exponential overhead of classical simulation dynamics will allow compilation larger algorithms, strategy this to evaluate an algorithm's cost on computer. To end, we propose variational hybrid quantum-classical algorithm called quantum-assisted compiling (QAQC). In QAQC, use overlap between target unitary $U$ trainable $V$ as function be evaluated More precisely, ensure QAQC scales well with problem size, our involves not only global ${\rm Tr} (V^\dagger U)$ but also local overlaps respect individual qubits. We introduce novel short-depth circuits quantify terms in function, prove cannot efficiently approximated under reasonable complexity assumptions. present gradient-free gradient-based approaches minimizing cost. As demonstration compile various one-qubit gates IBM's Rigetti's into their respective alphabets. Furthermore, successfully simulate up size 9 qubits, these simulations highlight scalability noise resilience QAQC. Future applications include depth compression, black-box compiling, mitigation, benchmarking.