Machine Learning for Scientific Analysis and Visualization

Abstract: In this talk, I will discuss some of my recent works on applying machine learning techniques for scientific data analysis and visualization. I will first give a brief overview of our previous work on constructing visualization surrogates using neural networks. The visualization surrogates allow quick previews of ensemble simulation results with freely selected simulation parameters. Then, I will talk about our more recent work on utilizing normalizing flows to perform super-resolution of scientific data. One unique advantage of using the normalizing flow model is to enable uncertainty analysis of the super-resolution result. Finally, I will discuss a method using scene representation networks (SRNs) to model three-dimensional scalar fields. We propose multi-grid SRN (APMGSRN) and employ a domain decomposition training and inference technique for accelerated parallel training on multi-GPU systems.

Bio: Han-Wei Shen is a Full Professor at The Ohio State University. He is the Editor-in-Chief of IEEE Transactions on Visualization and Computer Graphics, and an inductee of IEEE Visualization Academy. His primary research interests are scientific visualization and computer graphics. He received his BS degree from Department of Computer Science and Information Engineering at National Taiwan University, the MS degree in computer science from the State University of New York at Stony Brook, and the PhD degree in computer science from the University of Utah.