報告題目:Application of reinforcement learning to quantum control problems
報告人：王 欣 （香港城市大學物理系）
報告時間：2019年7月3日（星期三）上午9:30 — 11:00
In this talk, I will present results from our recent attempts to apply reinforcement learning techniques to quantum control problems. I will discuss how reinforcement learning can be used to generate controlling pulse sequences that prepare a desired quantum state, and its efficiency is compared to two other non-machine-learning methods, namely the stochastic gradient descent and Krotov methods . We have found that reinforcement learning has a particular advantage: it can adaptively optimize the sequences so that the simplest and most optimal ones are found. Moreover, reinforcement learning is also suitable for problems in which controls are discrete, while the Krotov method strongly favors continuous problems. I will also discuss our recent work to apply a version of reinforcement learning to quantum parameter estimation . We have shown that the control generated by our method is more transferrable than traditional methods such as GRAPE, namely the pulse sequences generated by the trained neuron network can be easily used to measure parameters having a range of values with reasonable precision. Our work shows that reinforcement learning can be a judicious numerical method when optimal quantum control is desired.
 X.-M. Zhang, Z. Wei, R. Asad, X.-C. Yang, and XW, arXiv:1902.02157.
 H. Xu, J. Li, L. Liu, Y. Wang, H. Yuan, and XW, arXiv:1904.11298.
Dr. Xin (Sunny) Wang received B.S. from School of Physics, Peking University in 2005, and received his Ph.D. degree from Columbia University in 2010. His Ph.D. study was focused on the theory of strongly correlated materials, in particular the high-Tc superconductors. From 2010-2015, Dr. Wang was a Research Associate in Condensed Matter Theory Center at University of Maryland, College Park. Starting from 2015 he joined City University of Hong Kong as an Assistant Professor. His current research interests include the theory of quantum computation using electron spins, correlated electron systems, and numerical methods. He has published 40 journal papers, including those in Nature Communications, npj Quantum Information, and Physical Review Letters.