Study of phase transition of Potts model with Domain Adversarial Neural Network
Xiangna Chen,Feiyi Liu,Feiyi Liu,Shiyang Chen,Jianmin Shen,Weibing Deng,Gábor Papp,Wei Li,Wei Li,Chunbin Yang
DOI: https://doi.org/10.1016/j.physa.2023.128666
2023-03-10
Abstract:A transfer learning method, Domain Adversarial Neural Network (DANN), is introduced to study the phase transition of two-dimensional q-state Potts model. With the DANN, we only need to choose a few labeled configurations automatically as input data, then the critical points can be obtained after training the algorithm. By an additional iterative process, the critical points can be captured to comparable accuracy to Monte Carlo simulations as we demonstrate it for q=3,4,5,7 and 10. The type of phase transition (first or second-order) is also determined at the same time. Meanwhile, for the second-order phase transition at q=3, we can calculate the critical exponent ν by data collapse. Furthermore, compared to the traditional supervised learning, we found the DANN to be more accurate with lower cost.
physics, multidisciplinary