ARU – towards automatic unfolding of detector effects

This article presents the ARU algorithm, a general non-interactive algorithmfor the unfolding of detector effects (resolution effects, efficiency, non-linearresponse) from one-dimensional data distributions. ARU uses an unbinnedmaximum-likelihood fit with a weighted regularization term, based on the rel-ative information in the solution with respect to a reference distribution. Theoptimal regularization weight is found by minimizing the mean squared errorof the solution. The algorithm’s performance is demonstrated in a study of atoy data sets. The analysis shows that the bias on average is smaller than thestatistical uncertainties which are properly estimated by the fit.