Speaker
Description
The diffuse supernova neutrino background (DSNB) describes the constant flux of neutrinos from past core-collapse supernovae over the visible universe.
The upcoming Jiangmen Underground Neutrino Observatory (JUNO), a 20 kton liquid scintillator detector, plans to observe the DSNB through the inverse beta decay (IBD) detection channel.
While other electron anti-neutrino sources will cause irreducible IBD background, non-IBD backgrounds such as neutron-induced events and NC interactions of atmospheric neutrinos can mimic the IBD event signature. We plan to reduce this background by machine-learning(ML)-based pulse-shape discrimination (PSD).
In this talk, we will compare three available ML techniques, feed-forward neural networks, support vector machines, and boosted decision trees, regarding both their achievable performance as well as the information these models provide about their given input parameters.
This work has been supported by the Clusters of Excellence PRISMA+ and ORIGINS as well as the DFG Collaborative Research Center "NDM" (SFB1258) and the DFG Research Units 2319 and 5519.
