18–20 Nov 2026
Bodoland University, Kokrajhar, Assam
Asia/Kolkata timezone

Application of machine learning techniques for model-independent transverse spherocity estimation in relativistic heavy-ion collisions

Not scheduled
20m
Bodoland University, Kokrajhar, Assam

Bodoland University, Kokrajhar, Assam

Debargaon, P.O. - Rangalikhata, J.D. Road, Kokrajhar(BTR), Assam, (India) Pin - 783370
Oral Presentation Track III: Computational Physics, Machine Learning & AI Applications in Physics

Speaker

Dipankar Basak (Kokrajhar University)

Description

In recent years, machine learning has been successfully used across many areas of high-energy physics. This study demonstrates the application of various machine learning (ML) algorithms to estimate the spherocity observable in heavy-ion collisions from experimentally measurable final-state observables. Transverse spherocity is an event-shape observable that effectively characterizes jet-like and isotropic event topologies in hadronic and heavy-ion collisions. We employed several supervised ML models optimized through systematic hyperparameter tuning to estimate both conventional and unweighted transverse spherocity in Au+Au collisions at $\sqrt{s_{\rm NN}}=200$ GeV using three experimentally measurable observables, namely charged-particle multiplicity at mid-rapidity $\langle dN_{\rm ch}/d\eta \rangle$, the average transverse mass of charged particles $\langle m_{\rm T} \rangle$, and the elliptic flow parameter $\langle v_2 \rangle$. The predictions from all the ML models agree well with the corresponding values obtained from AMPT simulations. Among the ML algorithms explored, the Light Gradient Boosting Machine (LightGBM) achieved the best overall performance, with unweighted spherocity consistently yielding lower MAE and RMSE and higher $R^2$ values than the conventional definition. Furthermore, the consistent performance across multiple event generators demonstrates the robustness of the ML approach against generator dependence, highlighting its potential for estimating spherocity from experimentally accessible observables in experimental data.

Authors

Dipankar Basak (Kokrajhar University) Kalyan Dey

Presentation materials

There are no materials yet.