Machine learning hadron spectral functions in Lattice QCD
The hadron spectral functions play an important role in understanding the properties of hadrons as well as thermodynamic properties of the QCD. For instance, the fate of quarkonia, which is considered as a thermometer of the hot medium, can be read-off from the distortion of their spectral functions in the thermal environment. However, the spectral functions are not directly accessible in the first-principle based lattice QCD simulations. They have to be reconstructed from the corresponding two-point correlation functions which are computable on the lattice. This reconstruction, however, is a typical ill-posed inverse problem. In this talk we present a novel neural network (sVAE) based on the variational auto-encoder and Bayesian theorem to study the inverse problem. As the number of solutions to the inverse problem is infinite, the proposed sVAE is designed to obtain the most probable image of the spectral function. We will discuss the mock tests using the sVAE and the application to charmonium correlation functions obtained from quenched lattice QCD simulations. The talk is based on arXiv:2110.13521.
Heng-Tong DingSpeakerProf. Heng-Tong Ding received his PhD in Bielefeld University, Germany in 2010. After that he went to Brookhaven National Laboratory and then Columbia University, USA as a postdoc researcher. He joined Central China Normal University (CCNU) as a professor in 2013, and is the vice director of the key laboratory of quark & lepton physics under the ministry of education of China at CCNU since 2019. His main research interest is to understand the QCD thermodynamics via the first-principle lattice QCD simulations. His publication records can be found in https://inspirehep.net/authors/1259382