论文标题
部分可观测时空混沌系统的无模型预测
Size optimization of CNOT circuits on NISQ
论文作者
论文摘要
量子计算机如今需要严格的内存约束，其中只能在图形结构中彼此之间的量子位之间执行2个Qubit的操作。因此，量子电路必须在实现图之前对图进行转换。在本文中，我们研究了一些嘈杂的中间量子量子（NISQ）设备上CNOT电路的优化。与以前的工作相比，我们将其分解为两个子问题：具有给定初始量子分布和优化的优化，而无需限制初始量子分布。我们发现，以前的大多数研究都集中在第一个子问题上，而忽略了相同拓扑结构中量子位不同分布对优化结果的影响。在本文中，我们考虑了两个子问题，并提供了一些新的优化算法。简而言之，我们的方法分为两个步骤：矩阵优化和路由优化。我们在[XZL+20]中使用算法实现矩阵优化，并使用MILP方法提出了一种新的启发式算法，该算法可以解决第二步。我们在IBM20和其他一些NISQ设备上实施算法，结果比我们实验中的大多数其他方法都要好。
Quantum computers in practice today require strict memory constraints, where 2-qubit operations can only be performed between the qubits closest to each other in a graph structure. So a quantum circuit must undergo a transformation to the graph before it can be implemented. In this paper, we study the optimization of the CNOT circuits on some noisy intermediate-scale quantum(NISQ) devices. Compared with previous works, we decompose it into two sub-problems: optimization with a given initial qubit distribution and optimization without limitations of initial qubit distribution. We find that most of the previous researches focused on the first sub-problem, and ignored the influence of different distribution of qubits in the same topology structure on the optimization results. In this paper, We take both sub-problems into account and give some new optimization algorithms. In short, our method is divided into two steps: matrix optimization and routing optimization. We implement matrix optimization with the algorithm in [XZL+20] and put forward a new heuristic algorithm with MILP method which can solve the second step. We implement our algorithm on IBM20 and some other NISQ devices, the results are better than most other methods in our experiment.
