Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2403.16389 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917653707227136 |
|---|---|
| author | Fox, Jordan M. R. Wendt, Kyle A. |
| author_facet | Fox, Jordan M. R. Wendt, Kyle A. |
| contents | We introduce a novel method for studying systematic trends in nuclear reaction data using generative adversarial networks. Libraries of nuclear cross section evaluations exhibit intricate systematic trends across the nuclear landscape, and predictive models capable of reproducing and analyzing these trends are valuable for many applications. We have developed a predictive model using deep generative adversarial networks to learn trends from the inelastic neutron scattering channel of TENDL for even-even nuclei. The system predicts cross sections based on adding/subtracting particles to/from the target nucleus. It can thus help identify cross sections that break from expected trends and predict beyond the limit of current experiments. Our model can produce good predictions for cross section curves for many nuclides, and it is most robust near the line of stability. We also create an ensemble of predictions to leverage different correlations and estimate model uncertainty. This research marks an important first step in computer generation of nuclear cross-section libraries. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_16389 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Illuminating Systematic Trends in Nuclear Data with Generative Machine Learning Models Fox, Jordan M. R. Wendt, Kyle A. Nuclear Theory Computational Physics We introduce a novel method for studying systematic trends in nuclear reaction data using generative adversarial networks. Libraries of nuclear cross section evaluations exhibit intricate systematic trends across the nuclear landscape, and predictive models capable of reproducing and analyzing these trends are valuable for many applications. We have developed a predictive model using deep generative adversarial networks to learn trends from the inelastic neutron scattering channel of TENDL for even-even nuclei. The system predicts cross sections based on adding/subtracting particles to/from the target nucleus. It can thus help identify cross sections that break from expected trends and predict beyond the limit of current experiments. Our model can produce good predictions for cross section curves for many nuclides, and it is most robust near the line of stability. We also create an ensemble of predictions to leverage different correlations and estimate model uncertainty. This research marks an important first step in computer generation of nuclear cross-section libraries. |
| title | Illuminating Systematic Trends in Nuclear Data with Generative Machine Learning Models |
| topic | Nuclear Theory Computational Physics |
| url | https://arxiv.org/abs/2403.16389 |