RACE: A Modular and Interpretable Radial–Angular Correlation Descriptor for Atomistic Machine Learning
6 affiliations
- 1University of Brasília, Institute of Physics, Brasília, Federal District 70910-900, Brazil
- 2Department of Physics, Federal University of Paraná, 81531-980 Curitiba, PR, Brazil
- 3Interdisciplinary Center for Science, Technology, and Innovation (CICTI), Federal University of Paraná, 81530-000 Curitiba, PR, Brazil
- 4Institute of Geosciences, University of Brasília, Brasília 70910-900, DF, Brazil
- 5Department of Physics, Norwegian University of Science and Technology (NTNU), NO-7491 Trondheim, Norway
- 6Computational Materials Laboratory, LCCMat, Institute of Physics, University of Brasília, 70910-900 Brasília, Federal District, Brazil
Abstract
Atomistic machine learning (ML) representations involve trade-offs among predictive accuracy, physical interpretability, descriptor dimensionality, and computational cost. Here, we introduce the Radial–Angular Correlation Descriptor (RACE), a modular representation constructed from radial distributions, angular distributions, and an explicit joint radial–angular map. Chemical identity is incorporated through interchangeable scalar or multichannel weighting functions rather than through a mandatory set of element-pair channels. We evaluate RACE against conventional smooth overlap of atomic positions (SOAP), compressed SOAP-μ2, atom-centered symmetry functions (ACSF), and a fixed-basis atomic cluster expansion (ACE) representation using common train–test partitions, training-only feature selection, repeated evaluations, and uncertainty estimates. RACE provides useful predictive accuracy across structurally and chemically diverse systems, but it does not uniformly outperform the alternative representations. Compressed SOAP is particularly accurate for the chemically heterogeneous benchmarks, whereas explicit composition features are required for robust generalization across broad Materials Project chemical spaces. Direct ablation of the radial, angular, and joint channels further shows that the incremental value of the joint radial–angular term is system dependent. The principal practical characteristics of RACE are therefore its transparent component structure, configurable chemical encoding, descriptor dimension independent of the number of species in its scalar-weighted form, and low peak memory during descriptor generation. RACE is consequently presented as a simple and manipulable structural representation that complements, rather than universally replaces, systematic density-correlation descriptors.
Related
- Descriptor-guided thermodynamic screening of H2 adsorption on single-atom-doped anatase TiO2 nanoparticles with interpretable machine learning
- An Efficient Descriptor for Tubular LDPE Melt Strength via Physics-Guided Learning and Diffusion-Based Inverse Design
- A Lightweight Universal Machine-Learning Interatomic Potential via Knowledge Distillation for Scalable Atomistic Simulations
- Overcoming the limitations of body-ordered potentials for atomistic machine learning
- From sparse quantum-computing data to atomistic simulation with universal machine-learning interatomic potentials
BibTeX
@article{tromer2026race,
title = {{RACE: A Modular and Interpretable Radial–Angular Correlation Descriptor for Atomistic Machine Learning}},
author = {Raphael M. Tromer and Jhionathan de Lima and Cristiano Francisco Woellner and Paola Ferreira Barbosa and Roberto Ventura Santos and Luiz Antonio Ribeiro Junior},
journal = {Journal of Chemical Theory and Computation},
year = {2026},
month = oct,
doi = {10.1021/acs.jctc.6c01111},
url = {https://doi.org/10.1021/acs.jctc.6c01111},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - RACE: A Modular and Interpretable Radial–Angular Correlation Descriptor for Atomistic Machine Learning
AU - Raphael M. Tromer
AU - Jhionathan de Lima
AU - Cristiano Francisco Woellner
AU - Paola Ferreira Barbosa
AU - Roberto Ventura Santos
AU - Luiz Antonio Ribeiro Junior
PY - 2026
DA - 2026/10/08
JO - Journal of Chemical Theory and Computation
DO - 10.1021/acs.jctc.6c01111
UR - https://doi.org/10.1021/acs.jctc.6c01111
AB - Atomistic machine learning (ML) representations involve trade-offs among predictive accuracy, physical interpretability, descriptor dimensionality, and computational cost. Here, we introduce the Radial–Angular Correlation Descriptor (RACE), a modular representation constructed from radial distributions, angular distributions, and an explicit joint radial–angular map. Chemical identity is incorporated through interchangeable scalar or multichannel weighting functions rather than through a mandatory set of element-pair channels. We evaluate RACE against conventional smooth overlap of atomic positions (SOAP), compressed SOAP-μ2, atom-centered symmetry functions (ACSF), and a fixed-basis atomic cluster expansion (ACE) representation using common train–test partitions, training-only feature selection, repeated evaluations, and uncertainty estimates. RACE provides useful predictive accuracy across structurally and chemically diverse systems, but it does not uniformly outperform the alternative representations. Compressed SOAP is particularly accurate for the chemically heterogeneous benchmarks, whereas explicit composition features are required for robust generalization across broad Materials Project chemical spaces. Direct ablation of the radial, angular, and joint channels further shows that the incremental value of the joint radial–angular term is system dependent. The principal practical characteristics of RACE are therefore its transparent component structure, configurable chemical encoding, descriptor dimension independent of the number of species in its scalar-weighted form, and low peak memory during descriptor generation. RACE is consequently presented as a simple and manipulable structural representation that complements, rather than universally replaces, systematic density-correlation descriptors.
SN - 1549-9618
SN - 1549-9626
ER -