NOvA interaction vertexing using a regression based Convolutional Neural Network

The NuMI Off-Axis $\nu_e$ Appearance (NOvA) experiment is an international neutrino oscillation experiment located at Fermi National Accelerator Laboratory (FNAL). NOvA observes interactions within two detectors – a Near Detector close to FNAL and a Far Detector at a distance of 810 km away in Ash River, Minnesota. These detectors, filled with liquid scintillator, collect scintillation light deposited by charged particles to form pixels. With this information, we reconstruct the energy and momentum of these particles to estimate the neutrino's energy, which is crucial to meaningfully constrain the oscillation parameters relevant to NOvA: $\Delta m_{32}^2$, $\sin^2 \theta_{23}$ and $\delta_{CP}$. To achieve this, an important component is estimating the location of the primary neutrino interaction vertex as accurately as possible. An incorrect vertex estimate can lead to an improper neutrino energy estimate, thus impacting NOvA's constraint on the oscillation parameters. This poster presents efforts to improve the estimation of the vertex in the detectors using a regression based Convolutional Neural Network. We analyze the performance of our network and present qualitative and quantitative metrics to assess our network over the traditional vertexing algorithms in NOvA.