AI Advances and Applications Virtual Seminar – Dr. Paul Torrey
- This event has passed.
AI Advances and Applications Virtual Seminar – Dr. Paul Torrey
October 21, 2020 @ 12:00 pm - 1:30 pm
AI Advances and Applications Virtual Seminar Series
“Machine Learning on the Cosmos” by Dr. Paul Torrey
Assistant Professor in Department of Astronomy
Wednesday, October 21, 2020
ABSTRACT:
Machine learning (ML) and artificial intelligence (AI) have an important role in interpreting the complex data that form an everyday part of modern astrophysics research. However, throughout all of astrophysics, we face a unique challenge in that we cannot directly access a ‘ground truth’ dataset that one would normally employ to train AI/ML methods. In the absence of such training sets, we must create them. Our research group builds, runs, and analyzes massive cosmological simulations as a way to probe the nature structure formation in our Universe. These same simulations can also be used as training sets for AI/ML analysis. In this talk, I’ll start by describing our traditional, computationally intensive approaches to modeling the Universe. I’ll then discuss two specific examples where simulation trained ML methods have come to play an important role by improving our ability to invert highly-complex, non-linear functions and back-out physical properties that would otherwise be nearly impossible to access. Finally, I’ll conclude by discussing some of the future prospects within the astrophysics and cosmology research spaces for applications of AI/ML analysis.
Bio:
Dr. Paul Torrey received his B.S. in Applied Physics from Cornell in 2008 followed by a Ph.D. in Astrophysics from Harvard in 2014. Prior to joining the UF faculty in 2018, Paul spent 4 years as a postdoc splitting his time between MIT and Caltech, including 2 years as a NASA Hubble Fellow. Paul’s research focus is on computational astrophysics with a focus on simulating the formation of structure in the Universe. He is a founding member of the Illustris collaboration, which has produced some of the largest and most accurate simulations of the Universe to date.