BIMSA >
BIMSA-Tsinghua Seminar on Machine Learning and Differential Equations
Data driven method to learn the stochastic dynamical systems and its application in polymer dynamics
Data driven method to learn the stochastic dynamical systems and its application in polymer dynamics
Organizers
Speaker
Xiaoli Chen
Time
Thursday, November 10, 2022 9:00 AM - 10:30 AM
Venue
1129B
Online
Zoom 537 192 5549
(BIMSA)
Abstract
In this talk, I will discuss how to use machine learning method to learn the stochastic problem. To begin, I will introduce the how to combine physics informed neural network(PINN) method and the sample observation data to learn the stochastic differential equation driven by Brown and Levy noise. Second, I will introduce how to use stochatic OnsagerNet to learn closure dynamical systems. We propose a general machine learning approach to construct reduced models for noisy, dissipative dynamics based on the Onsager principle for non-equilibrium systems. Then I will demonstrate our method by modelling the folding and unfolding of a long polymer chain in an external field - a classical problem in polymer rheology - though our model is suitable for the description of a wide array of complex, dissipative dynamical systems arising in scientific and technological applications.
Speaker Intro
陈小丽，2020年博士毕业于华中科技大学。2018年9月至2020年8月在美国布朗大学进行联合培养。2021年3月至今在新加坡国立大学数学系和功能智能材料研究院(I-FIM)做博士后研究。主要从事随机动力系统, 机器学习与动力系统的研究。已在SIAM Journal on Scientific Computing, Physica D, Chaos等期刊发表多篇学术论文。