Upcoming High Energy Theory, Relativity, and Cosmology Seminars


What can Machine Learning teach us about phase transitions? -- Biagio Lucini

Jul 26, 2019, 12:45 PM-1:30 PM

202 Physics Bldg.

Host: Simon Catterall/ Contact: Yudaisy Salomón Sargentón, 315-443-5960

Recently, there has been a surge of studies of phase transitions using machine learning techniques. Most of the works so far have focused on qualitative or semi-quantitative features. In this talk, I will present an investigation of the Ising model using an interpretable Machine Learning method, the Support Vector Machine. Acting solely on Monte Carlo generated data and without any assumption on the underlying Hamiltonian, I shall show that over the same set of configurations the method provides quantitative results that are comparable in precision with those obtained through a standard analysis based on finite size scaling of reweighed data. I will also discuss the lesson learnt from this study for the reconstruction of the relevant symmetry driving the phase transition.