Hit Position Reconstruction with Machine Learning Techniques for TPEX

The two-photon exchange (TPE) contribution in elastic electron-proton scattering has been of significant interest to the nuclear physics community as a possible explanation for the proton form factor ratio discrepancy. Three experiments (VEPP-3, CLAS12, and OLYMPUS) have reported on direct measurements of TPE in recent years and show effects in agreement with phenomenological predictions, but in disagreement with theory, up to $Q^2$ = 2.5 (GeV/c)$^2$. At larger $Q^2$, for which other theoretical approaches, like those based on GPDs, become feasible the effect remains untested. The proposed Two-Photon Exchange eXperiment (TPEX) will measure hard two-photon contribution to $ep$ scattering in the momentum transfer range up to $Q^2$ = 4.7 (GeV/c)$^2$. The experiment uses electromagnetic calorimeter blocks at fixed angles to detect scattered electrons. In this talk we will discuss using a Deep Neural Net (DNN) and Multi-Layer Perceptron (MLP) to reconstruct the $ep$ hit position, and compare to the simple analytical solution. Some prelimary results from a Monte Carlo simulation of the experiment will be shown, and future work will be discussed.