The Project 8 collaboration is developing Cyclotron Radiation Emission Spectroscopy (CRES) on the tritium $\beta$-decay spectrum for measurement of the absolute neutrino mass, aiming for a final mass target sensitivity of 0.04 eV/$c^2$. The general principle of CRES experiments is to build an energy spectrum by reconstructing the start frequencies of quasi-linear charged particle trajectories (called tracks) in frequency and time. In Project 8, neutrino mass inference relies on a well-understood and precise reconstruction of the $\beta$-spectrum endpoint region where a variation due to the presence of a massive neutrino is maximal. Due to the relatively small number of events in this region and the need for excellent instrumental energy resolution, reconstruction methods which are both efficient and highly accurate are desired. Particularly suited to deal with information-sparse data and offering highly-accurate foreground localization via segmentation, deep learning convolutional neural networks (CNNs) may be used to extract track properties from CRES events with relative ease. In this work, we develop a novel machine-learning based model which operates a CNN and a support vector machine in tandem to reconstruct simulated CRES events. We also show improvements in efficiency and accuracy of event parameters when compared to a traditional point-clustering based approach.
This work is supported by the US DOE Office of Nuclear Physics, the US NSF, the PRISMA+ Cluster of Excellence at the University of Mainz, and internal investments at all institutions.