Mini-symposium (MS) 

MS9:  Deep Learning Methods for Data Driven Models

Organizers: Zhu Wang, University of South Carolina

                     Lili Ju, University of South Carolina

Abstract: Modern sensor technology and data acquisition capability have led to the explosive production of digital data and information. Data-driven approaches based on model reduction methods or machine learning techniques are powerful tools for extracting important characteristics and essential representations from the massive data, affecting every branch of science and social life with unprecedented impact. Although deep learning has achieved tremendous successes in many areas such as computer vision and speech recognition, challenges still exist in many scientific and engineering areas. This mini-symposium will focus on recent advances in data-driven approaches using emerging deep learning algorithms together with their applications in scientific research and engineering.

Saturday, September 18, 10:00 AM – 12:00 PM: Part I of II

Room: Libry 3127

10:00 – 10:30 Wuchen Li, University of South Carolina, Alternating the population and control neural networks to solve high-dimensional stochastic mean-field Games

10:30 – 11:00 Shu Liu, Georgia Tech, Neural parametric Fokker-Planck equations

11:00 – 11:30 Wei Zhu, University of Massachusetts Amherst, Equivariant neural network in computer vision and scientific computing

11:30 – 12:00 Jia Zhao, Utah State University, Solving and learning phase-field models using the modified physics informed neural networks

Saturday, September 18, 3:30 PM – 5:30 PM: Part II of II

Room: Libry 3127

3:30 – 4:00 Chunmei Wang, University of Florida, Reproducing activation functions for deep learning

4:00 – 4:30 Senwei Liang, Purdue University, Solving PDEs on unknown manifolds with machine learning

4:30 – 5:00 Guannan Zhang, Oak Ridge National Laboratory, A nonlocal gradient for high-dimensional black-box optimization in scientific machine learning

5:00 – 5:30 Anthony Gruber, Florida State University, Convolutional neural networks for data compression and reduced order modeling