报告题目：Accelerate calculations of transition starts by GPR

报告人：周翔副教授, 香港城市大学

报告时间：2020年11月6日15:00-16:00

腾讯会议：931 916 378；密码：1122

报告摘要：The saddle point (SP) detecting is a grand challenge for computational intensive energy function. The traditional method usually be time-consuming due to the thousands simulation for the derivative information of the energy function. We propose a surrogate based SP detection method which obviously reduce the number of simulations. This method combines a statistical method, Gaussian process regression and a dynamic SP detecting method, Gentle accent dynamic method. We sequentially detecting the SP by GAD and update the surrogate GPR. We use an active learning strategy to detect the points getting data and a quick numerical implementation algorithm is given. Two classical rare event examples are given for comparing the efficiency of the method. A challenge MD example are given.

报告人简介: 周翔, 香港城市大学数据科学学院和数学学院副教授、博士生导师。博士毕业于普林斯顿大学，曾在布朗大学做博后研究工作。在J. Comp. Phys. SIAM J. Numer. Anal. SIAM J. Sci. Comput. Nonlinearity和 J. Chem. Phys.等期刊发表SCI论文10余篇。