Collaborative Research: OAC Core: An Integrated Framework for Enabling Temporal-Reliable Quantum Learning on NISQ-era Devices

随着量子计算机不断扩大规模，增加量子比特，利用量子计算开发实际和现实世界的应用程序已成为量子信息科学家和技术人员的重要前沿，这使其他科学家和多个学科的最终用户受益。量子学习是量子计算和机器学习的结合，也被称为变分量子算法(VQA)，是解决各种实际问题的最有前途的方法之一。量子学习是一种混合量子-经典协议，它使用经典的训练优化器来优化具有代价函数的变分量子电路(VQC)中的参数。然而，量子器件固有的噪声带来了严重的可部署性和便携性问题，使得优化后的VQC在不同量子计算机之间部署或移植时性能显著下降。更重要的是，量子器件上的噪声随着时间的推移而变化，称为不稳定噪声、波动噪声或噪声漂移，这阻碍了在不同时间在一台量子计算机上重复使用VQC，甚至在VQC训练过程中当噪声变化时将学习误导到非最优路径。该项目旨在通过在量子学习、不确定性预测、噪声抑制和系统可视化方面产生基本的理解和实用的方法，使时间可靠的量子学习成为可能。在能源部赞助的超级计算中心使用量子学习进行科学应用程序的结果评估，这些中心提供对各种商业量子计算资源的访问。本项目以促进实际量子学习为目标，采用系统创新的方法开发了一个集成的框架，展示了所提出的研究的新颖性、实用价值和领域影响：(1)开发了一种新型的基于压缩的误差适配器来根据波动的量子噪声调整VQC的参数和结构，使VQC能够有效和高效地适应当前的量子噪声；(2)构建不确定性预测器来量化给定的VQC和量子处理器的可部署性，从而使用户能够知道性能的变化；(3)设计一个具有可扩展性的新型可视化工具来描述噪声对给定VQC性能的影响；以及(4)所开发的工具集最终集成到一个科学应用程序--实时地震检测中，它可以为识别现实世界中的任务提供洞察力，其中量子技术可能提供一个有前途的解决方案。该项目的教育影响包括有关开发的软件工具的教程，以指导和鼓励领域研究人员利用先进的量子计算网络基础设施；将波托马克量子创新中心针对高中生的研究整合到量子暑期计划中；以及为量子劳动力培训开发新的本科生和研究生课程。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

As quantum computers consistently scale up with more qubits, the development of practical and real-world applications using quantum computing has become a crucial frontier for quantum information scientists and technologists, which benefits other scientists and end-users across a wide range of disciplines. Quantum learning, a combination of quantum computing and machine learning and also known as Variational Quantum Algorithm (VQA), is one of the most promising approaches to be applied to a variety of practical problems. Quantum learning is a hybrid quantum-classical protocol that optimizes parameters in a Variational Quantum Circuit (VQC) with a cost function using a classical training optimizer. However, the inherent noise on quantum devices brings severe deployability and portability issues, making the optimized VQCs suffer significant performance degradation in deploying or porting among different quantum computers. What is more, the noise on the quantum devices changes over time, known as unstable noise, fluctuating noise, or drift of noise, which prevents the reuse of VQCs on one quantum computer at different times and even misleads the learning to a non-optimal path when noise change during the VQC training process. This project aims to enable temporal-reliable quantum learning by generating fundamental understandings and practical approaches in quantum learning, uncertainty prediction, noise suppression, and system visualization. Outcomes are evaluated using quantum learning for scientific applications on the DoE-sponsored supercomputing centers that provide access to various commercial quantum computing resources. With the objective of facilitating practical quantum learning, this project uses a systematic and innovative approach to develop an integrated framework, which presents the novelty of the proposed research, practical value, and domain impacts: (1) developing a novel compression-based error adaptor to adjust the parameters and structure of VQC according to the fluctuating quantum noise, such that the VQC can effectively and efficiently adapt to the present quantum noise; (2) building an uncertainty predictor to quantify the deployability of a given pair of VQC and quantum processor, such that users can be aware of performance change; (3) designing a novel visualization tool with scalability to portray the impact of noise on the performance of a given VQC; and (4) the developed toolset is finally integrated into a scientific application, real-time earthquake detection, which can provide insights into identifying real-world tasks where quantum technologies may offer a promising solution. The education impacts of this project include the tutorials on the developed software tools to guide and encourage the domain researchers to leverage the advanced quantum computing cyberinfrastructure; the integration of research to the quantum summer program from the Potomac Quantum Innovation Center for high school seniors; and the development of new undergraduate and graduate courses for quantum workforce training.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.