Past Event: Oden Institute Seminar

Machine Learning and Astronomy

Josh Peek, Space Telescope Institute

Abstract

The use of machine learning in astronomy has exploded over the last 5 years. In this talk I will define “machine learning” for our context, and talk about where I think machine learning can be useful for astronomy. I will review the history of machine learning in astronomy and touch on some favorite use cases. I’ll assess where we are and where I think we are going. I will dive into two particular astronomical use cases that my group has been exploring. The first is how machine learning on Galaxy morphology can inform our understanding of the physics of galaxy formation and evolution. The second is an unsupervised learning approach to the “search by image” problem in astronomy, and some interesting overlaps with citizen science.

Biography

Josh Peek is the Head of Data Science and Astronomical Archives at the Space Telescope Science Institute, as well as a associate astronomer with tenure. Josh did undergrad work at Harvard, a brief stint at the Submillimeter Array, and PhD work at UC Berkeley in astrophysics where he focused on radio astronomy and the diffuse interstellar medium of the Milky Way. As a postdoctoral research and Hubble Fellow at Columbia he expanded his work into nearby galaxies, interstellar and intergalactic dust, the galaxy’s circumgalactic medium, and machine vision and statistical techniques for complex astronomical images. At STScI he is part of the world’s largest research group devoted to the study of the interstellar medium and has focused more on applications of ML and AI to astronomical and astroinformatics problems.