29 May 2018 to 3 June 2018
Hyatt Regency Indian Wells Conference Center
US/Pacific timezone

The Application of Deep Learning to Event-by-Event Simulations of Relativistic Hydrodynamics

31 May 2018, 18:10
20m
North Foyer | Ironwood Room (Hyatt Regency Indian Wells Conference Center)

North Foyer | Ironwood Room

Hyatt Regency Indian Wells Conference Center

44600 Indian Wells Lane, Indian Wells, CA 92210, USA

Speaker

Dr LongGang Pang (UC Berkeley and LBNL)

Description

The state-of-the-art pattern recognition method in machine learning (deep convolution neural network) has been used to classify two different phase transitions between normal nuclear matter and hot-dense quark gluon plasma. Large amounts of training data have been prepared by simulating heavy ion collisions with event-by-event relativistic hydrodynamics. High level correlations of particle spectra in transverse momentum and azimuthal angle learned by the neural network are quite robust in deciphering the transition type in the quantum chromodynamics phase diagram. Through this study, we demonstrated that there is a traceable encoder of the phase structure that survives the dynamical evolution and exists in the final snapshot of heavy ion collisions and one can exclusively and effectively decode this information from the highly complex output using machine learning.
E-mail lgpang.1984@berkeley.edu

Primary author

Dr LongGang Pang (UC Berkeley and LBNL)

Co-authors

Presentation materials