Composable and adaptive design of machine learning interatomic potentials guided by Fisher-information analysis
This work addresses the challenge of creating flexible and extensible interatomic potentials for materials science, representing an incremental improvement in model design methods.
The paper tackled the problem of designing machine-learning interatomic potentials by proposing an adaptive strategy that iteratively reconfigures composite models guided by Fisher-information analysis and error metrics, resulting in an optimal configuration with 75 parameters achieving a force RMSE of 0.172 eV/Å and an energy RMSE of 0.013 eV/atom.
An adaptive physics-informed model design strategy for machine-learning interatomic potentials (MLIPs) is proposed. This strategy follows an iterative reconfiguration of composite models from single-term models, followed by a unified training procedure. A model evaluation method based on the Fisher information matrix (FIM) and multiple-property error metrics is proposed to guide model reconfiguration and hyperparameter optimization. Combining the model reconfiguration and the model evaluation subroutines, we provide an adaptive MLIP design strategy that balances flexibility and extensibility. In a case study of designing models against a structurally diverse niobium dataset, we managed to obtain an optimal configuration with 75 parameters generated by our framework that achieved a force RMSE of 0.172 eV/Å and an energy RMSE of 0.013 eV/atom.