Campus, Varian 355
The nature and volume of observed and simulated astronomical data present unique challenges to artificial intelligence/machine learning (AI/ML) and visualization methods. In this talk, I will present three new AI/ML and visualization techniques motivated by the needs of astronomical research that enable us to better understand and interpret the universe. I will first survey some novel methods for source detection, deblending, and morphological classification that leverage recent advances in computer vision. I’ll then discuss our work creating FitsMap, a new, performant tool for displaying large volumes of astronomical image and catalog data that scales to large volumes of data. I will conclude by showing how an interpretable ML model called Explainable Boosting Machines can illuminate the relationships between galaxy star formation rate, stellar mass, halo properties, and environment. The success of these methods support the view that AI/ML has an important and increasing role in deepening our understanding of complex astrophysics.