Speaker
Description
A densely connected feed-forward neural network is capable to classify poles of scattering amplitude if fed with experimentally measured values of energy-dependent production intensity. As shown in [1], such a neural network trained with synthetic intensities based on effective range approximated amplitudes classifies the $P_c(4312)$ signal as a virtual state located at the 4th Riemann sheet in momentum space with very high certainty. This is in line with the results of other analyses but surpasses them by providing the simultaneous evaluation of probabilities of competing scenarios, like eg. the interpretation as a bound state. The machine learning approach also allows for identifying the energy bins which are key for the physical interpretation.
Here we discuss the extended approach, where the neural network is used to classify the physical nature of not just one state but the whole class of states described by the coupled channel amplitudes dominated by a pole lying close to the threshold, like $Σ_c^+\bar{D}^0$ threshold for $P_c(4312)$ or $K\bar{K}$ threshold for $a_0(980)$ or $f_0(980)$ resonances. Apart from the fundamental significance for interpreting the nature of various hadron candidates, our approach has a practical application for experimental analyses by providing the means of rapid classification of potentially exotic states.
Bibliography
1. Deep Learning Exotic Hadrons, JPAC Collaboration • L. Ng (Florida State U.) et al., e-Print: 2110.13742 [hep-ph]
| In-person participation | Yes |
|---|
