Speakers
Description
The Fermilab Muon g-2 experiment aims to measure the magnetic anomaly of the muon at the precision level of 140 ppb. The data model to extract the anomalous precession frequency is subjected to various beam dynamics corrections, contributing significantly to the systematics of the extracted frequency. These beam dynamics corrections were estimated by monte-carlo simulations, which often are not computationally efficient, given the target statistics of the experiment. Modeling these effects using machine learning-based models can provide a more efficient solution to this problem. We explored various machine learning models (such as Boosted Decision Tree, PDEFoam, Artificial Neural Network, and K-nearest neighbors) of the acceptance of positron events in the calorimeters. The models' performances are compared, and the BDT and PDEFoam models can achieve an Area Under the Receiver Operation Curve (AUROC) of over 0.8.
