Please visit Jefferson Lab Event Policies and Guidance before planning your next event: https://www.jlab.org/conference_planning.

May 8 – 12, 2023
Norfolk Waterside Marriott
US/Eastern timezone

Application of Machine Learning to Particle Identification at the BESIII experiment

Not scheduled
1h
Hampton Roads Ballroom and Foyer Area (Norfolk Waterside Marriott)

Hampton Roads Ballroom and Foyer Area

Norfolk Waterside Marriott

235 East Main Street Norfolk, VA 23510
Poster Poster Poster Session

Speaker

Chen, Zhengyuan (Institute of High Energy Physics Chinese Academy of Sciences)

Description

Particle identification is an important ingredient to particle physics experiments. Distinguishing the charged hadrons (pions, kaons, protons and their antiparticles) is often crucial, in particular for hadronic decays which could be studied with an efficient particle identification to obtain a desirable signal-to-background ratio. An optimal performance of particle identification in a large momentum range requires an effective combination of various relevant variables provided by almost all sub-systems of a general-purpose detector. Since the particle identification capability of each variable has a complicated dependence on particle momentum and potential correlations between the other variables, it is intricate to obtain a perfect performance utilizing a predetermined equation as a model. Machine learning algorithms have been developing to use computational methods to “learn” information directly from data. Particle identification is a typical application of classification techniques of machine learning which involves a large amount of data and lots of features. The BESIII detector is used for studies of hadron physics and $\tau-$charm physics, which is composed of a helium-gas based drift chamber, a time-of-flight system, a CsI(Tl) crystal electromagnetic calorimeter and a resistive plate chamber based muon counter. High statistics and high purity data samples of pion, kaon and (anti-)proton are selected using real data accumulated in BESIII experiment, and the detector responses of these hadron samples have been investigated and summarized. These hadron samples and their characteristics of detector responses offer an unique opportunity to take advantages of machine learning techniques to make classifications or predictions. With application of gradient boosted decision trees (BDT) and usage of various features provided by all sub-detectors, the inherent potentialities of BESIII detector for particle identification is exploited and an enhancement of hadron identification capability has been achieved, especially for high momentum particles.

Consider for long presentation Yes

Primary authors

Chen, Zhengyuan (Institute of High Energy Physics Chinese Academy of Sciences) Dr Sun, Shengsen (Institute of High Energy Physics Chinese Academy of Sciences)

Co-authors

Presentation materials