AccelNet: Exact backward-compatible acceleration of polynomial angular descriptors through Cartesian moment factorization
Abstract
We present AccelNet, an exact, backward-compatible method for accelerating existing trained aenet and n2p2 neural-network potentials without retraining. For angular terms with separable one-neighbor weights and a finite polynomial dependence on cos θ, the method exploits their hidden finite-rank structure to replace explicit neighbor-pair loops by one-neighbor Cartesian moments. AccelNet reads models trained with either package and reproduces their descriptors, energies, and analytic forces to floating-point roundoff. We verified this equivalence for H2O and TiO2 models and tested the resulting potentials in LAMMPS molecular-dynamics simulations. The implementation, model-conversion tools, and LAMMPS interfaces are released as open-source software. AccelNet also enables GPU-accelerated molecular dynamics with existing aenet and n2p2 potentials. Moment evaluation achieves speedups of up to 10.9 on a CPU and 15.8 on a GPU relative to direct evaluation within AccelNet. The implementation, model-conversion tools, and LAMMPS interfaces are released as open-source software.
Related
- Cartesian isotropic tensors revisited
- Fascia anisotropic damage propagation with microstructure- level descriptors
- Physics-guided deep neural networks for Laguerre–Gaussian modal decomposition via PCA-ranked Zernike descriptors
- Reduced vortex descriptors linking polycrystallinity in magnetic nanoparticles with polarized magnetic small-angle neutron scattering
- GEODE: Symmetry-Preserving Cartesian Diffusion for Crystal Generation
BibTeX
@misc{nagai2026accelnet,
title = {{AccelNet: Exact backward-compatible acceleration of polynomial angular descriptors through Cartesian moment factorization}},
author = {Yuki Nagai},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2608.03280},
eprint = {2608.03280},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.03280},
}
RIS
TY - UNPB
TI - AccelNet: Exact backward-compatible acceleration of polynomial angular descriptors through Cartesian moment factorization
AU - Yuki Nagai
PY - 2026
DA - 2026/09/30
PB - arXiv
DO - 10.48550/arxiv.2608.03280
UR - https://arxiv.org/abs/2608.03280
AB - We present AccelNet, an exact, backward-compatible method for accelerating existing trained aenet and n2p2 neural-network potentials without retraining. For angular terms with separable one-neighbor weights and a finite polynomial dependence on cos θ, the method exploits their hidden finite-rank structure to replace explicit neighbor-pair loops by one-neighbor Cartesian moments. AccelNet reads models trained with either package and reproduces their descriptors, energies, and analytic forces to floating-point roundoff. We verified this equivalence for H_2O and TiO_2 models and tested the resulting potentials in LAMMPS molecular-dynamics simulations. The implementation, model-conversion tools, and LAMMPS interfaces are released as open-source software. AccelNet also enables GPU-accelerated molecular dynamics with existing aenet and n2p2 potentials. Moment evaluation achieves speedups of up to 10.9 on a CPU and 15.8 on a GPU relative to direct evaluation within AccelNet. The implementation, model-conversion tools, and LAMMPS interfaces are released as open-source software.
ER -