报告题目：Quantum reservoir computing and its potential applications

报告人：李晓鹏 教授（复旦大学）

报告时间：2025年10月31日（周五）上午10:30

报告地点：光电所二层报告厅

报告摘要：  

Reservoir computing provides a promising route to temporal machine learning by exploiting the intrinsic nonlinear dynamics of a complex physical system while only training an inexpensive readout. Quantum reservoir computing (QRC) extends this idea to quantum many-body dynamics, offering access to richer state spaces and new memory–nonlinearity tradeoffs. I will first summarize theoretical work that characterizes reservoir learning power across the quantum many-body-localized (MBL) to ergodic transition, showing that optimal performance often lies at the edge between memory retention and nonlinearity. Next, I will introduce a practical engineering approach — configured QRC — which uses gradient-free optimization to program NISQ devices so a single quantum reservoir can learn multiple, heterogeneous tasks with accuracy beyond classical reservoirs. Finally, I will present recent experimental progress in small correlated-spin reservoirs that achieves state-of-the-art temporal predictions, illustrating how few-qubit quantum systems can outperform much larger classical reservoirs on real tasks. I will close with perspectives on device platforms, task selection, and avenues for scaling QRC toward integrated quantum-machine-learning applications.

报告人简介：

李晓鹏，复旦大学物理系教授。2008年本科毕业于中国科学技术大学，2013年在美国匹兹堡大学获得博士学位。2013-2016年在马里兰大学从事博士后研究，2016底加入复旦大学物理系任青年研究员，2017年入选海外高层次人才计划，同年入选福布斯中国U30科技精英榜，2019年晋升正教授，2020年起在上海期智研究院兼任杰出科学家，2021年获上海市青年科技启明星计划资助，2022年入选复旦大学青年谢希德教授，2024年获得上海青年五四奖章。主要从事量子计算和量子模拟的研究，近期专注于基于中性原子的容错量子计算机研制。