Speaker
Description
The Multi-Purpose Detector (MPD) experiment is a flagship heavy-ion experiment of the NICA facility at JINR, in Dubna expected to start operation in 2026. The experiment will operate in the energy range $\sqrt{s_{\rm NN}}$ = 4-11 GeV in collider mode and $\sqrt{s_{\rm NN}}$ = 2.4-3.5 GeV in fixed-target mode which covers the high net-baryon density region of the QCD phase diagram.
Particle identification plays a critical role in experimental data analysis, in particular, electron identification which is essential for measuring dileptons in heavy-ion collisions. Traditional methods based on one-dimensional selection criteria often suffer from loss of efficiency due to multiple sequential selection cuts on discriminating variables. In this presentation, we discuss the application of machine learning techniques to enhance electron identification in the MPD experiment. Classifiers were trained on electron samples, and their performance was evaluated and compared to that of the conventional cut-based approach. The results demonstrate the potential of machine learning to significantly improve electron identification efficiency while maintaining an almost 100% purity.
