Data Science Techniques for Systems with a Sign Problem

The sign problem stands in the way of advancing our understanding of many quantum materials, making usual numerical methods to study them impossible. This work uses machine learning and other data science techniques to compare different methods of circumventing the sign problem in many-body quantum systems. Data visualization techniques and machine learning have been useful in understanding these stochastic methods, and in identifying bottlenecks and numerical issues in the algorithms. We examine data from both Monte Carlo simulations with analytical continuation and from complex Langevin simulations in order to improve our approaches to studying systems with a sign problem.