Machine learning methods for track classification in the AT-TPC

Abstract

We evaluate machine learning methods for event classification in the Active-Target Time Projection Chamber detector at the National Superconducting Cyclotron Laboratory (NSCL) at Michigan State University. Currently, events of interest are selected via cuts in the track fitting stage of the analysis workflow. An explicit classification step to single out the desired reaction product would result in more accurate physics results as well as a faster analysis process. We tested binary and multi-class classification methods on data produced by the 46 Ar(p,p) experiment run at the NSCL in September 2015. We found that fine-tuning a pre-trained convolutional neural network produced the most successful classifier of proton scattering events in the experimental data, when trained on both experimental and simulated data. We present results from this investigation and conclude with recommendations for event classification in future experiments.

Digital Signal Analysis based on Convolutional Neural Networks for Active Target Time Projection Chambers

“…For instance, CNNs have shown a great performance for event classification 2 [16], pulse shape discrimination, [17], pileup correction [18], event classification [19] and many more [20]. CNNs have the ability to fit multi-variate functions and to learn complex mathematical procedures from a large set of trainable parameters.…”

Classification of events from $α$-induced reactions in the MUSIC detector via statistical and ML methods

“…For example, in the case of the Multi-Sampling Ionization Chamber (MUSIC) detector data, the number of background events is typically several orders of magnitude more than the events of interest, rendering out-of-the-box anomaly detection methods ineffective. A few recent examples of domain-specific AI/ML methods can be found in the works of Kuchera et al [1] (experiment) and Raghavan et al [2] (theory). For a recent review, see [3].…”

“…Our proposed approach comprises four phases: (1) filter easy-to-identify background events; (2) detect and remove background events that are similar to the events of interest; (3) get input from the domain scientists and design a classifier; and (4) detect the strip location at which the event took place. We leverage statistical and machine learning methods in phases 1-3.…”

Performance tests of a LaBr3:Ce detector coupled to a SiPM array and the GET electronics for γ-ray spectroscopy in a strong magnetic field

“…Among these active gas target projects, there are systems where the detection gas is put inside of solenoid magnets of B = 2-4 T [9,10,11,12,13]. The radius-of-curvature of charged particles within these magnetic fields may be used to measure their energies [10,14,15], thus enabling for particle detection over a dynamic range larger than that of silicon semiconductor detectors, which are typically used for charged particle detection in nuclear physics investigations. In the case of the SpecMAT detector (being built at KU Leuven), γ-ray detection by means of scintillation crystals is added in conjunction with an active target in a strong magnetic field, thus providing a powerful tool for nuclear spectroscopy.…”