Machine learning approaches for giant dipole resonance parameters

One of the basic types of nuclear collective excitations is giant dipole resonance (GDR), which has energy greater than the binding energy per nucleon. The study of GDR can contribute to the understanding of nuclear structure. The experimental GDR observables have been measured in various types of experiments, namely, photonuclear experiments with photons from Bremsstrahlung radiation, and more recently, laser Compton scattering (LCS), as well as (p, p') reaction, and so on. Experimental data of most nuclei near the β-stability line are available in major photonuclear data libraries (RIPL, IAEA, CENDL, etc.). Our ML models learned and predicted critical GDR parameters, such as peak energies and resonance widths more accurately than the phenomenological GT and SJ models. As an improvement to the present work, the determination of a complete and global set of GDR parameters using ML technique is under progress