Active-subspace learning reveals dominant parameter directions in nuclear observables
Abstract
Nuclear observables computed using ab initio methods can depend on many low-energy constants (LECs) of the underlying chiral effective field theory interactions, while being sensitive to only a few LEC combinations. We use active-subspace learning to identify these dominant parameter directions directly from many-body calculations. For ground-state energies and charge radii of oxygen isotopes, we find that more than 99% of the gradient-based variation is concentrated along a single direction for each observable in the 17-dimensional LEC space. Comparing these directions across nuclei and observables quantifies similarities in their LEC dependence. The same analysis yields derivative-based global sensitivities without requiring a global emulator. Active-subspace directions can also serve as an efficient representation for emulation, greatly improving Gaussian-process emulators especially for broad LEC variations. We demonstrate this for Hartree-Fock calculations of 48Ca, where the emulator error is improved with active-subspace learning by more than two orders of magnitude.
Related
- Hamiltonian-Guided Leverage Embedding: Robust Subspace Compression for Efficient QAOA Parameter Estimation
- Exploring new directions in enhancing the ACTS parameter optimization suite
- Learning to Program Adaptive Non-Local Observables for Machine Learning
- Learning in the Transverse Subspace: A Minimal Representation for Divergence-Free Operator Learning
- Machine Learning β-decay Half-lives and Their Application to r-Process Observables
BibTeX
@misc{plies2026activesubspace,
title = {{Active-subspace learning reveals dominant parameter directions in nuclear observables}},
author = {Tom Plies and Margarida Companys Franzke and Achim Schwenk},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.03628},
eprint = {2611.03628},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.03628},
}
RIS
TY - UNPB
TI - Active-subspace learning reveals dominant parameter directions in nuclear observables
AU - Tom Plies
AU - Margarida Companys Franzke
AU - Achim Schwenk
PY - 2026
DA - 2026/11/02
PB - arXiv
DO - 10.48550/arxiv.2611.03628
UR - https://arxiv.org/abs/2611.03628
AB - Nuclear observables computed using ab initio methods can depend on many low-energy constants (LECs) of the underlying chiral effective field theory interactions, while being sensitive to only a few LEC combinations. We use active-subspace learning to identify these dominant parameter directions directly from many-body calculations. For ground-state energies and charge radii of oxygen isotopes, we find that more than 99% of the gradient-based variation is concentrated along a single direction for each observable in the 17-dimensional LEC space. Comparing these directions across nuclei and observables quantifies similarities in their LEC dependence. The same analysis yields derivative-based global sensitivities without requiring a global emulator. Active-subspace directions can also serve as an efficient representation for emulation, greatly improving Gaussian-process emulators especially for broad LEC variations. We demonstrate this for Hartree-Fock calculations of ^48Ca, where the emulator error is improved with active-subspace learning by more than two orders of magnitude.
ER -