Speaker
Description
Jets have been used as an indirect probe to study the properties of the quark gluon plasma (QGP). Machine learning (ML) models are a useful means for performing studies in general because they can learn complex patterns and relationships from data without requiring explicit analytic solutions. This is particularly useful when it is difficult to obtain solutions using traditional methods, as is the case with modeling of QGP related effects. To extend our understanding of the path length dependence of jet modifications, we developed techniques using ML optimized for making physically constrained predictions. This study uses experimentally measurable jet observables to predict jet quenching with extensions planned for parton flavor tagging and characterizing background. In this talk, we will present a score, the “Z-metric” which can provide levels of significance of these observables for predicting the amount of quenching experienced by the jet. Pythia, Jewel and Hydjet event generators are used in training samples to assess if the predictions made are model independent. This study utilizes a gradient boosted random forest model, although the presented metric can be applied to neural networks of various types. The model interpretability, made possible by the proposed Z-metric, allows for the isolation of different physics processes (soft radiation background, flow, energy transfer to generated QGP medium, dispersion in QGP medium) present in the measured observables by decoupling the input features, making this a promising tool for experimental analysis. These developments will be used to characterize jets in experimental data.
| Is this an experimental talk? | Yes |
|---|---|
| Is this on behalf of a collaboration? | No |
| Are you willing to present as a poster if it is not selected for oral presentation? | Yes |