Speaker
Description
Understanding the non-perturbative confinement regime of quantum chromodynamics (QCD) necessitates identifying states within the hadronic spectrum. Recently, numerous new states have been discovered by various experimental collaborations. However, not all observed signals correspond to excitations of low-lying hadrons. Rigorous amplitude analysis techniques are essential to determine which observed signals are genuinely part of the hadronic spectrum.
In this talk, I will discuss how machine learning can be utilized in amplitude analysis to classify observed signals. Specifically, a deep neural network (DNN) can be trained to map input line shape space to output interpretation space. To ensure the DNN functions as a universal approximator, the training dataset must consist of input line shapes generated from a model-independent general amplitude parametrization. I will demonstrate that a trained DNN can distinguish the nature of poles using only the line shape above the threshold of a single-channel two-hadron scattering system. Additionally, I will show that kinematical enhancements, such as the triangle singularity, can be differentiated from pole-based enhancements using the same DNN principles.
