CRA Seminar: Yuan-Sen Ting (ANU)

Title: Transcending the Limits of Theoretical Physics with Deep Learning

Abstract: Astronomy has recently experienced a profound transformation, driven by the acquisition of extensive datasets generated by increasingly sophisticated instruments. This influx of data has opened up numerous new avenues for exploration in the field. However, this boon is accompanied by its own set of challenges, as astronomical phenomena often exhibit intricate and are  inherently high-dimensional. These complexities involve precise imaging, spectral analysis, and time series data.
Traditional statistical methods in astronomy can struggle to effectively handle these complexities. Deep learning offers a solution by effectively addressing the challenges by allowing for a more faithful representation of the intricate astronomical phenomena. In this talk, I will delve into deep learning approaches for characterizing complex astronomical systems. Additionally, it will explore the theoretical underpinnings of deep learning, including its close relationship with symmetry and physics. This exploration will encompass a diverse range of applications in astronomy, including asteroseismology, stellar spectroscopy, weak lensing, reionization, galactic dynamics, galaxy evolution, and quasars. These applications collectively contribute to expanding the boundaries of Bayesian statistics and theoretical astrophysics, all within the framework of deep learning.