Thursday, December 19, 20192:00 PM - 3:00 PMCNLS Conference Room (TA-3, Bldg 1690)|
Learning Nuclear Masses with Mixture Density Networks
Amy LovellT-CNLS, T-2
The mass of each nucleus is a basic quantity but gives rise to many other properties of nuclear structure and reactions, including binding energies, separation energies, and Q-values. Close to stability, masses can be measured extremely precisely through various experimental techniques. Farther away from stability, in regions of the nuclear chart necessary for many studies including for fission or astrophysics, masses must be calculated. There have been several recent studies using various statistical techniques to correct these mass models based on available experimental data and to provide quantified uncertainties that can be propagated to other quantities of interest. Here, we expand this type of calculation, using a rather novel machine learning technique, the Mixture Density Network, to directly predict the nuclear masses with uncertainties. We show preliminary calculation in the context of previous machine learning methods and then discuss future work.
Host: Tillmann Weisser