Resonance Searches using Gaussian Process Regression in the Heavy Photon Search Experiment

The Heavy Photon Search (HPS) experiment in Jefferson Lab’s Hall B searches for a kinetically mixed vector boson A′ (“dark photon”) produced via electroproduction in fixed-target electron-nucleus scattering and decaying to electron-positron pairs. In the regime where the A′ decays are prompt (i.e., at larger ε²), the signal appears in the electron-positron invariant mass distribution as a narrow, Gaussian-like excess on top of the smooth QED background. Traditional functional-form background modeling approaches are challenged by HPS’s complex geometric acceptance and high-statistics, motivating development of Gaussian-process regression (GPR) analysis techniques. This presentation covers HPS-specific GPR advances including adaptive sideband selection and alpha dependent noise modeling. Additional results include signal injection/extraction studies, background validation, and an expected sensitivity reach estimate for the 2021 (160 pb^-1 ) dataset. Preliminary findings from the GPR analysis on the 2015 (1.2 pb^-1) and 2016 (10 pb^-1) engineering run datasets, as well as a randomized fractional subset of the 2021 data, will also be presented.