AIMS Home

Scientific machine learning for dynamics related inverse problems

Organizer(s):
Name:	Affiliation:	Country:
Yanhzao Cao	Auburn University	USA
Feng Bao	Florida State University	USA
Guannan Zhang	Oak Ridge National Lab	USA

Abstract:

Scientific Machine Learning (SciML) has recently received significant attention in the research communities that involve large-scale data and complex models. Successfully trained neural networks, enabled by massive computing power and colossal amount of data, have led to quantum leaps in artificial intelligence. Deep learning techniques are positioned to fundamentally change many sectors of society, by offering decision making capabilities which match and often exceed that of human experts. The power of SciML is not limited to learning forward models. It is also a powerful tool to solve the inverse problem, which is an important area in scientific research that aims to combine forward simulation with data to build-up inferences for target scientific models. In this special session we focus on the application of SciML on solving scientific inverse problems, especially the dynamical system itself given input data, and we bring researchers together that have developed methods and algorithms on solving inverse problems with help of state-of-the-art SciML techniques. The topics of interest include but not limited to learning dynamical systems, stochastic optimization, Bayesian inference, data assimilation, probabilistic machine learning, and scientific data analytics.

List of approved abstract
Name	Affiliation	Title
Rick Archibald	Oak Ridge National Laboratory	Federated Learning for Scientific Facilities
Feng Bao	Florida State University	A Stochastic Maximum Principle Approach for Reinforcement Learning
Yanzhao Cao	Auburn University	On a stochastic neural network
Wuchen Li	University of South Carolina	Controlling regularized conservation laws via entropy-entropy flux pairs
MIroslav Stoyanov	Oak Ridge National Lab	Multidimensional Quadrature Rule for Non-Positive Weight Functions
Hoang A Tran	Oak Ridge National Laboratory	High-Dimensional Optimization with a Novel Nonlocal Gradient
Hans Werner Van Wyk	Auburn University	Incorporating auxiliary data in parameter estimation through machine learning
Lijin Wang	University of Chinese Academy of Sciences	On structure-preserving numerical methods for Stochastic Poisson systems
Minglei Yang	Oak Ridge National Laboratory	A pseudo-reversible normalizing flow for stochastic dynamical systems with arbitrary initial distributions