VAQEM: A Variational Approach to Quantum Error Mitigation
VAQEM: A Variational Approach to Quantum Error Mitigation
复制标题
DOI：
10.1109/hpca53966.2022.00029
复制
发表时间：
2021-12
期刊：
影响因子：
--
通讯作者：
Gokul Subramanian Ravi;Kaitlin N. Smith;P. Gokhale;A. Mari;N. Earnest;Ali Javadi-Abhari;F. Chong
中科院分区：
文献类型：
--
作者：
Gokul Subramanian Ravi;Kaitlin N. Smith;P. Gokhale;A. Mari;N. Earnest;Ali Javadi-Abhari;F. Chong
变分量子算法（VQA）是近期量子优势的最有前途的候选者之一。传统上，这些算法通过旋转门角度来参数化，其值在量子机器上的迭代执行中进行调整。这些门角度参数的迭代调整使VQA对量子机的噪声分布更鲁棒。然而，噪声的影响仍然是VQA在真实的量子机器上的目标估计的一个重大损害-它们远非理想。因此，有必要采用有效的错误缓解策略，以提高这些量子算法的保真度在近期machines.While现有的错误缓解技术建立在理论上确实提供了大量的收益，理论和真实的机器执行特性之间的脱节限制了这些改进的范围。因此，它是至关重要的，以优化缓解技术，以明确地适应目标应用程序以及目标machine.We的噪声特性，它动态地调整现有的错误缓解技术的实际，动态的噪声执行特性的VQA目标量子机。我们这样做是通过调整这些缓解技术的具体功能类似于传统的旋转角度参数-通过针对一个特定的目标函数，代表手头的VQA问题的改进。在本文中，我们的目标是两种类型的错误缓解技术，适用于量子电路中的空闲时间：单量子比特门调度和插入动态解耦序列。我们获得了实质性的改进VQA客观测量-超过3倍的各种VQA应用程序的平均值，运行在IBM Quantum machines.More重要的是，虽然我们研究了两个特定的错误缓解技术，提出的变分方法是通用的，可以扩展到许多其他错误缓解技术的具体配置是很难选择先验。未来将更多的缓解技术集成到VAQEM框架中可以带来进一步的巨大收益，可能会在当今嘈杂的量子机器上实现实际有用的VQA优势。
Variational Quantum Algorithms (VQA) are one of the most promising candidates for near-term quantum advantage. Traditionally, these algorithms are parameterized by rotational gate angles whose values are tuned over iterative execution on quantum machines. The iterative tuning of these gate angle parameters make VQAs more robust to a quantum machine’s noise profile. However, the effect of noise is still a significant detriment to VQA’s target estimations on real quantum machines — they are far from ideal. Thus, it is imperative to employ effective error mitigation strategies to improve the fidelity of these quantum algorithms on near-term machines.While existing error mitigation techniques built from theory do provide substantial gains, the disconnect between theory and real machine execution characteristics limit the scope of these improvements. Thus, it is critical to optimize mitigation techniques to explicitly suit the target application as well as the noise characteristics of the target machine.We propose VAQEM, which dynamically tailors existing error mitigation techniques to the actual, dynamic noisy execution characteristics of VQAs on a target quantum machine. We do so by tuning specific features of these mitigation techniques similar to the traditional rotation angle parameters -by targeting improvements towards a specific objective function which represents the VQA problem at hand. In this paper, we target two types of error mitigation techniques which are suited to idle times in quantum circuits: single qubit gate scheduling and the insertion of dynamical decoupling sequences. We gain substantial improvements to VQA objective measurements — a mean of over 3x across a variety of VQA applications, run on IBM Quantum machines.More importantly, while we study two specific error mitigation techniques, the proposed variational approach is general and can be extended to many other error mitigation techniques whose specific configurations are hard to select a priori. Integrating more mitigation techniques into the VAQEM framework in the future can lead to further formidable gains, potentially realizing practically useful VQA benefits on today’s noisy quantum machines.