Seminar: Can Machine Learning Inspire Quantum Algorithms
Jiaqi Leng
Assistant Professor of Computer Science
Virginia Tech
Friday, October 23, 2026
2:30 - 3:30p.m.
3100 Torgersen Hall
Abstract
Quantum machine learning often asks whether quantum computers can accelerate learning. This talk asks the reverse question: can the fundamental ideas behind machine learning inspire new quantum algorithms? I will explore this question through two influential ideas in modern machine learning: gradient descent and score matching. First, inspired by the connection between accelerated gradient descent and Hamiltonian dynamics, we develop a quantum optimization framework called quantum Hamiltonian descent (QHD). By exploiting quantum dynamics, QHD offers new mechanisms for navigating nonconvex optimization landscapes, with provable super-polynomial speedups over classical algorithms for certain structured problems. Second, I will discuss how score matching, a fundamental technique behind modern diffusion models, can be extended from classical probability distributions to quantum states. This leads to efficient algorithms for learning quantum Hamiltonians from experimental data, achieving near-optimal sample complexity in certain parameter regime. Together, these examples illustrate a broader approach to quantum algorithm design: translating the mathematical principles behind successful machine learning methods into quantum dynamics and algorithms.
Biography
Jiaqi Leng is an Assistant Professor of Computer Science at Virginia Tech, based in the greater Washington, D.C. Area. He is also a Commonwealth Cyber Initiative (CCI) Faculty Fellow, and an affiliate of the Lawrence Berkeley National Lab. His research interest lies at the intersection of quantum computation, machine learning, and optimization. In particular, he aims to design efficient, practical, and scalable quantum algorithms for real-world applications.