应兰州大学核科学与技术学院、稀有同位素前沿科学中心邀请，深圳国际量子研究院吴步娇副研究员将于2026年10月10日来校进行学术交流并作报告。

报告题目： Quantum Learning and Variational Algorithms for Quantum Many-Body Problems

报告时间：2026年10月10日（星期六）16 : 30

报告地点：榆中校区秦岭堂A114

欢迎广大师生参加！

【报告摘要】

Quantum phase recognition and ground-state computation provide important settings for exploring the power of quantum computing, while posing shared challenges in algorithmic expressivity, hardware implementation, and trainability. This talk brings together studies of quantum learning and variational quantum algorithms to examine these challenges. For quantum phase recognition, quantum kernel methods learn phase structures characterized by linear order-parameter observables, with theoretical guarantees of efficiency and robustness. Under appropriate computational complexity assumptions, these methods reveal quantum learning advantages for a class of phase-recognition tasks, supported by numerical studies of symmetry-protected topological and symmetry-broken phases.[1] For ground-state computation, combining classical tensor networks with parameterized quantum circuits enhances the expressivity of variational states without increasing the physical circuit depth, with numerical demonstrations in quantum spin models.[2] On superconducting hardware, a proof-of-principle implementation of a measurement-based variational quantum eigensolver (MBQC-VQE) uses cluster-state resources to estimate ground-state energies of a perturbed planar-code model, demonstrating an experimental route to measurement-driven variational computation.[3] To further understand the capabilities and limitations of VQE circuits, we examine the mechanisms underlying barren plateaus. By distinguishing observable concentration from the loss of parameter sensitivity within the circuit, this analysis shows that avoiding observable concentration alone does not guarantee trainability: information loss or information scrambling can still produce exponentially suppressed gradients.[4] Together, these studies connect algorithm design, hardware demonstrations, and trainability analysis, providing insight into balancing expressivity and optimization efficiency in practical algorithms for quantum many-body problems under limited quantum resources.

 References

[1] Y. Wu, B. Wu, J. Wang, and X. Yuan. Quantum Phase Recognition via Quantum Kernel Methods. Quantum 7, 981 (2023). doi: 10.22331/q-2023-04-17-981.

[2] J. Huang, W. He, Y. Zhang, Y. Wu, B. Wu, and X. Yuan. Tensor-network-assisted variational quantum algorithm. Physical Review A 108(5), 052407 (2023). doi: 10.1103/PhysRevA.108.052407.

[3] S. Cao, B. Wu, F. Chen, et al. Generation of genuine entanglement up to 51 superconducting qubits. Nature 619(7971), 738–742 (2023). doi: 10.1038/s41586-023-06195-1.

[4] Z.-S. Li, B. Wu, X.-W. Li, and M.-H. Yung. Barren Plateaus Beyond Observable Concentration. arXiv preprint arXiv:2603.18479 (2026).

【报告人简介】

吴步娇，深圳国际量子研究院副研究员。2021年获中国科学院计算技术研究所计算机科学与技术博士学位，曾在北京大学和德国柏林自由大学从事博士后研究，并在南方科技大学任副研究员。主要研究方向为量子随机测量、量子学习理论、量子优越性算法设计与量子线路优化。

核科学与技术学院

稀有同位素前沿科学中心

2026年10月9日