报告题目：Machine Learning for Frustrated Classical Spin Models
报 告 人：王策博士 清华大学高等研究院
报告摘要：In this talk, I will show how to apply the machine learning method to study classical XY model on frustrated lattices, such as triangle lattice and Union-Jack lattice. The low temperature phases of these frustrated models exhibit both U(1) and Z2 chiral symmetry breaking, and therefore they are characterized by two order parameters, and consequently, two successive phase transitions as lowering the temperature. By using classical Monte Carlo to generate a large number of data to feed computer, I use methods such as the principle component analysis (PCA) to analyze these data. I will show that the PCA method can distinguish all different phases and locate phase transitions, without prior knowledge of order parameters. This method offers promise for using machine learning techniques to study sophisticated statistical models, and our results can be further improved by using principle component analysis with kernel tricks and the neural network method.
报告人简介:Ce Wang is now a PhD student of Institute for Advanced Study, Tsinghua University. He graduated from Department of Physics, Tsinghua University in 2014. His research field includes machine learning and cold atom physics.