A study of the scaling behavior of the two-dimensional Ising model by methods of machine learning

In the field of condensed matter physics, machine learning methods have become an increasingly important instrument for researching phase transitions. Here we present a method for calculating the universal characteristics of spin models using an Ising model that is exactly solvable in two dimensions. The method is based on a convolutional neural network (CNN) with controlled learning. The scaling functions prove the continuing type of phase transition for the 2D Ising model. As a result of the proposed technique, it has been possible to calculate correlation length directly.