Speaker
Description
The Compressed Baryonic Matter (CBM) experiment is an upcoming fixed target experiment being built at the Facility for Anti-proton and Ion Research (FAIR).
The CBM experiment is designed to characterize the QCD medium at high net baryon densities and moderate temperatures.
Di-electrons interact electromagnetically and are unaffected by strong medium effects.
Hence, they are used as a penetrating probe to understand the QCD medium produced in the initial stages of heavy-ion collisions.
The identification of electrons with minimal pion contamination is crucial for these kinds of investigations.
The CBM experiment uses a Ring Imaging Cherenkov detector (RICH) in combination with a Transition Radiation Detector (TRD) for electron-pion separation, and a Time of Flight
(TOF) detector for identification of other high-mass hadrons.
The current RICH reconstruction algorithm employs an Artificial Neural Network (ANN) as a conventional electron identifier, which utilizes ring and track parameters as inputs.
In a recent upgrade, a new XGBoost model was implemented, adding additional input features, replacing the conventional ANN.
The output of the RICH XGBoost model serves as a probability measure for selecting electrons.
Similarly, TRD and TOF have their own measures for the electron identification.
Currently, a global electron identifier using the information from RICH, TRD, and TOF is developed using tree-based ensemble models.
A momentum independent training strategy is used to train the model, as it is foreseen to work for different collision energies (Au-Au systems up to 11 AGeV beam energy) and centralities.
This contribution will focus on the design aspects and performance analysis of the global electron identifier model with corresponding feature optimization.
$^*$This work is supported by BMBF (05P21PXFC1, 05P24PX1).
| Category | Experiment |
|---|---|
| Collaboration (if applicable) | CBM |
