Graduate Student Seminar with Jason Bub on Uncertainty Quantification in Effective Field Theories

In the past few decades, nuclear physics has entered into an era of high precision and high accuracy, both in experiment and computational methods. Alongside these, we have seen the rise of complex microscopic nuclear interactions derived from effective field theories (EFTs), which yield high-quality results when paired with state-of-the-art computational methods. However, one crucial area has lagged with these strides: uncertainty quantification in theoretical calculations. Without robust methods of accounting for theoretical error, we are essentially lost on how to compare theoretical works to their experimental counterparts. Our approach to rectify this is to introduce a Bayesian framework when fitting models for the nuclear interactions, baking statistics into every step of the process. Doing so presents an intuitive way to quantify various sources of uncertainty in our calculations, from statistical to theoretical. However, implementing such an approach is not without its struggles, ranging from computational costs to numerical instability, all of which must be surmounted along the way through old and new methods, such as gaussian processes, machine learning, and surrogate models.