Exploring Thermodynamic Behavior of Spin Glasses with Machine Learning

In this paper, we consider the regression problem of predicting the average energy $\langle E \rangle$ as a function of temperature $T$ for spin glasses on a square lattice using several machine learning methods. Representing the spin glass as a weighted graph with exchange interactions as edges, we explored the relationship between the spatial distribution of connections and $\langle E \rangle$ using them.