Critical Temperature from Unsupervised Deep Learning Autoencoders

Andreas Athenodorou
2022 Zenodo  
We discuss deep learning autoencoders for the unsupervised recognition of phase transitions in physical systems formulated on a lattice. We elaborate on the applicability and limitations of this deep learning model in terms of extracting the relevant physics. Although our results are shown in the context of 2D, 3D and 4D Ising models as well as 2D Potts model, we focus on the analysis of the critical quantities at 2D (anti)ferromagnetic Ising Model. We define as a quasi-order parameter the
more » ... ute average latent variable, which enables us to predict the critical temperature. We demonstrate that we can define a latent susceptibility and use it to quantify the value of the critical temperature Tc(L) at different lattice sizes and that these values suffer from smaller finite scaling effects compared to what one obtains from the magnetic susceptibility.
doi:10.5281/zenodo.6405399 fatcat:bcqaehqn55ed7jliqrzicrcq6y