报告题目: Identifying QCD transition using deep learning
报 告 人: Yi-Lun Du, Nanjing University & Frankfurt Institute for Advanced Studies & Goethe Universität Frankfurt
时 间： 1月 8日（星期一）下午 15:30~16:30
地 点：3-323
报告摘要：
Deep learning (DL), as a branch of machine learning, can capture high-order correlations from big data. Recently, this technique has made impressive progresses in physics research, such as in particle physics and condensed matter physics. Great advantage of DL over conventional methods has revealed in the respect of extracting hidden features from highly dynamical and complex non-linear systems.
In this talk, I will discuss the application of DL in high energy nuclear physics. Supervised learning with a deep convolutional neural network (CNN) is used to identify the QCD equation of state (EoS) employed in event-by-event (2+1)-D relativistic viscous hydrodynamics coupled to a hadronic cascade "afterburner" simulations of heavy-ion collisions from the simulated final-state pion spectra $/rho(p_T, /phi)$. High-level correlations of $/rho(p_T,/phi)$ are learned by the neural network, which acts as an effective "EoS-meter" in distinguishing the nature of the QCD transition. The EoS-meter is robust against many simulation inputs, which provides a powerful tool as the direct connection of heavy-ion collision observables with the bulk properties of QCD.
