Uncertainty-Driven Self-Refinement of Adaptive Kolmogorov-Arnold Networks for Fitting Nonadiabatic Potential Energy Surface
Abstract
In this work, we propose an uncertainty-driven self-refinement learning method using adaptive Chebyshev-polynomial-based Kolmogorov-Arnold networks (ChebyKANs) to build potential energy surface (PES). Unlike conventional approaches, our method employs a main-model-plus-residual-model architecture for automatic sampling based on a hybrid criterion, where sampling and fitting are decoupled. To benchmark our method, the non-adiabatic two-state PES of the H + H2 system is constructed using a database computed at the multireference configuration interaction (MRCI) level with a rather large active space. Our calculations show that the proposed architectural design makes the efficiency gain, and that KAN offers enhanced interpretability over conventional neural network (NN) fitting methods through its mathematically well-understood basis functions. Furthermore, the present calculations accurately reproduce all important conformations on the PES, including the transition state and the conical intersection. Then, the multi-configurational time-dependent Hartree (MCTDH) method reproduces the non-adiabatic dynamics with results in close agreement with previously reported findings.
Related
- Pre-Refinement Prediction of Dopant Site Occupancy in M-Type Hexaferrites via Hierarchical and Uncertainty-Aware Machine-Learning.
- Continuous perception and adaptive metrics for scalable structure refinement
- AdaptNTK: Adaptive Uncertainty Quantification and Active Learning for Neural Network Potentials
- Active-Space Selection in Nonadiabatic Dynamics Simulations of Dense Manifolds of Excited States: When Local Diabatization Needs Great Care
- Adaptive Expansions of the Optimized Effective Potential in Physically Motivated Response Spaces
BibTeX
@misc{yu2026uncertaintydriven,
title = {{Uncertainty-Driven Self-Refinement of Adaptive Kolmogorov-Arnold Networks for Fitting Nonadiabatic Potential Energy Surface}},
author = {Jinke Yu and Xingyu Zhang and Qingyong Meng},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.38227},
eprint = {2610.38227},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.38227},
}
RIS
TY - UNPB
TI - Uncertainty-Driven Self-Refinement of Adaptive Kolmogorov-Arnold Networks for Fitting Nonadiabatic Potential Energy Surface
AU - Jinke Yu
AU - Xingyu Zhang
AU - Qingyong Meng
PY - 2026
DA - 2026/10/28
PB - arXiv
DO - 10.48550/arxiv.2610.38227
UR - https://arxiv.org/abs/2610.38227
AB - In this work, we propose an uncertainty-driven self-refinement learning method using adaptive Chebyshev-polynomial-based Kolmogorov-Arnold networks (ChebyKANs) to build potential energy surface (PES). Unlike conventional approaches, our method employs a main-model-plus-residual-model architecture for automatic sampling based on a hybrid criterion, where sampling and fitting are decoupled. To benchmark our method, the non-adiabatic two-state PES of the H + H_2 system is constructed using a database computed at the multireference configuration interaction (MRCI) level with a rather large active space. Our calculations show that the proposed architectural design makes the efficiency gain, and that KAN offers enhanced interpretability over conventional neural network (NN) fitting methods through its mathematically well-understood basis functions. Furthermore, the present calculations accurately reproduce all important conformations on the PES, including the transition state and the conical intersection. Then, the multi-configurational time-dependent Hartree (MCTDH) method reproduces the non-adiabatic dynamics with results in close agreement with previously reported findings.
ER -