InfoCGAN classification of 2D square Ising configurations

Abstract

An infoCGAN neural network is trained on 2-dimensional square Ising configurations conditioned on the external applied magnetic field and the temperature. The network is composed of two main sub-networks. The network learns to generate convincing Ising configurations as well as discriminate between ``real and ``fake configurations with an additional categorical assignment prediction. The predicted categorical assignments show very strong agreement with the expected physical phases in the Ising model, the ferromagnetic spin-up and spin down phases as well as the paramagnetic phase. Additionally, configurations associated with the crossover region in the non-vanishing field case are predicted by the model. The classification probabilities allow for a robust method of estimating the critical temperature in the vanishing field case, showing exceptional agreement with the known physics. This work indicates that an adversarial neural network approach can be used efficiently to identify physical phases with no a priori information beyond raw physical configurations and the physical conditions they are subject to.