Machine Learning Methods for Track Classification in the AT-TPC

Track classification methods were evaluated for reaction products in the Active-Target Time Projection Chamber (AT-TPC) at the National Superconducting Cyclotron Laboratory (NSCL). Single-class, binary, and multi-class event classification methods were benchmarked on data produced by the $^{46}$Ar(p,p) experiment which ran at the NSCL in September of 2015. The experiment ran with a 1.68 T magnetic field parallel to the beam axis, producing spiral tracks. Results from logistic regression, feed-forward neural networks, convolutional neural networks, and support vector machines will be presented. We will compare classification of experimental data with simulated training data versus experimental training data. Recommendations for choosing appropriate event classification methods in future AT-TPC experiments will be made.