Hugo Sträng: Applying Machine Learning to Number-Theoretical Data

Abstract

We explore number-theoretical problems using machine learning techniques, specifically focusing on class groups of imaginary quadratic fields. We investigate methods for predicting divisibility properties of class numbers based on discriminant information; however, our models show poor performance in these tasks. Additionally, we implement methods for approximating the number of negative fundamental prime discriminants with class groups of a given order. In this task, our models achieve relatively low errors.