Electron-Nucleus Cross Sections from Transfer Learning.

Area of Science:

• Nuclear Physics
• Computational Physics
• Machine Learning

Background:

• Transfer learning is a machine learning technique where a model trained on one task is repurposed for a second related task.
• Deep neural networks (DNNs) have shown promise in various scientific domains.

Purpose of the Study:

• To investigate the application of transfer learning using DNNs in nuclear physics.
• To assess the efficacy of fine-tuning DNNs trained on one scattering process for predicting outcomes of related processes.

Main Methods:

• Training deep neural networks on inclusive electron-carbon scattering data.
• Fine-tuning the trained DNNs for new prediction tasks.
• Validating predictions against experimental cross-section data for various nuclear targets.

Main Results:

• The fine-tuned DNNs accurately predicted cross sections for electron interactions with nuclear targets.
• Successful application of transfer learning across a range of nuclei from helium-3 to iron.

Conclusions:

• Transfer learning is a viable and effective technique for accelerating predictions in nuclear physics.
• DNNs can be efficiently adapted to model diverse electron-nucleus scattering processes.