NuCLR: Nuclear Co-Learned Representations
This work addresses the challenge of accurately modeling nuclear properties for fundamental physics research, offering potential insights into nuclear theory.
The authors tackled the problem of predicting nuclear observables like binding energies and charge radii by introducing NuCLR, a deep learning model using multi-task training with shared representations, achieving state-of-the-art precision crucial for nuclear astrophysics. They also found that the learned representation captures key aspects of the nuclear shell model, including magic numbers and the Pauli Exclusion Principle.
We introduce Nuclear Co-Learned Representations (NuCLR), a deep learning model that predicts various nuclear observables, including binding and decay energies, and nuclear charge radii. The model is trained using a multi-task approach with shared representations and obtains state-of-the-art performance, achieving levels of precision that are crucial for understanding fundamental phenomena in nuclear (astro)physics. We also report an intriguing finding that the learned representation of NuCLR exhibits the prominent emergence of crucial aspects of the nuclear shell model, namely the shell structure, including the well-known magic numbers, and the Pauli Exclusion Principle. This suggests that the model is capable of capturing the underlying physical principles and that our approach has the potential to offer valuable insights into nuclear theory.