Transfer Learning Approaches Leveraging Nuclear Ab Initio Reaction Models

This PhD project aims to enhance the predictive accuracy of nuclear models by integrating machine learning with ab initio methods for nuclear reactions. The project focuses on developing Artificial Neural Networks (ANNs) to predict nuclear cross-sections, using pseudo-scattering data generated from toy-model Hamiltonians. The pretrained ANNs will be fine-tuned using transfer learning, adapting to experimental data or ab initio reaction calculations, to extend their applicability to complex nuclear collisions beyond few-body systems. The goal is to provide systematic evaluations of nuclear reactions, incorporating multiple reaction channels, and addressing data limitations in nuclear astrophysics. The project will also explore advanced many-body techniques like Configuration Interaction (CI), Resonating Group Method (RGM), and No-Core Shell Model with Continuum (NCSMC) to improve calculation accuracy, with potential applications in nuclear, hadronic, and atomic physics.