Electron-nucleus cross sections from transfer learning

2024-08-19

arXiv.org

AI Summary

Traditional electron–nucleus cross-section modeling requires separate model development and training for each nuclide, hindering generalizability and scalability. Method: We propose the first transfer learning framework for nuclear cross-section prediction, pretraining a deep neural network on electron–carbon scattering data and fine-tuning it with minimal target-nuclide data (¹² nuclides from ³He to ⁵⁶Fe). Crucially, no architectural modification or de novo training is needed. Contribution/Results: The approach overcomes the long-standing limitation of target-specific model reuse in nuclear physics. After fine-tuning on small datasets, it achieves sub-5% mean relative error across light-to-medium-mass nuclides (e.g., Li–Fe), substantially outperforming classical parametric models. This work pioneers the systematic application of transfer learning to data-driven nuclear cross-section modeling, establishing a scalable, multi-nuclide, few-shot paradigm for nuclear reaction modeling.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Transfer learning (TL) allows a deep neural network (DNN) trained on one type of data to be adapted for new problems with limited information. We propose to use the TL technique in physics. The DNN learns the physics of one process, and after fine-tuning, it makes predictions for related processes. We consider the DNNs, trained on inclusive electron-carbon scattering data, and show that after fine-tuning, they accurately predict cross sections for electron interactions with nuclear targets ranging from lithium to iron. The method works even when the DNN is fine-tuned on a small dataset.

Problem

• Predict electron-nucleus cross sections using transfer learning
• Adapt DNN trained on electron-carbon data to other nuclei
• Improve accuracy for cross sections from helium-3 to iron

Innovation

• Transfer learning adapts DNN for new physics problems
• DNN trained on electron-carbon data predicts cross sections
• Fine-tuning enables accurate predictions for various nuclei