Speaker
Description
The Mu2e experiment will search for the neutrino-less conversion of muons to electrons in muonic aluminum. The Mu2e tracker detector measures the momentum of signal and background particles traveling down the beamline. One major background source is the $\mu\rightarrow e \bar{\nu} \nu$ decay of muons in orbit (DIO) process. Due to resolution and algorithm errors during reconstruction, these electrons can become background in the signal momentum selection window. A track quality selection algorithm using TMVA to filter out low-quality tracks and most of the DIO electrons was previously described in Using Machine Learning to Select High-Quality Measurements by Andrew Edmonds et al. In this presentation we present an alternate track quality selection algorithm using a stochastic gradient-boosted decision tree in Python's scikit-learn library. We improved momentum track quality training time and classification time by a factor of four when tested on the Mu2e tracker simulation data.
| Consider for long presentation | No |
|---|
