Partial Differential Equations & Applied Math Seminar
Dr. Ming Zhong, University of Houston
We develop a learning framework to perform system identification for understanding the observation data of a single snapshot from observing collective behaviors. Such learning problem can become extremely ill-conditioned. Our proposed learning provides enough regularization to tackle such ill-conditioning. These tools include the incorporation of the empirical distribution of static patterns. We investigate special scenarios of such data, including steady states and scenarios of patterns remaining relatively steady. We provide extensive examples to show the efficiency and effectiveness of our learning. This is a joint work with Baoli Hao (Illinois Tech) and Mauro Maggioni (Johns Hopkins).