DOE OSTI · 2999427
When more data hurts: Optimizing data coverage while mitigating diversity-induced underfitting in an ultrafast machine-learned potential
Abstract
Machine-learned interatomic potentials (MLIPs) are becoming an essential tool in materials modeling. However, optimizing the generation of training data used to parametrize the MLIPs remains a significant challenge. This is because MLIPs can fail when encountering local environments too different from those present in the training data. The difficulty of determining a priori the environments that will be encountered during molecular dynamics simulation necessitates diverse, high-quality training data. Here, this study investigates how training data diversity affects the performance of MLIPs using the Ultra-Fast force field (UF 3 ) to model amorphous silicon nitride. We employ expert and autonomously generated data to create the training data and fit four force field variants to subsets of the data. Our findings reveal a critical balance in training data diversity: insufficient diversity hinders generalization, while excessive diversity can exceed the MLIP's learning capacity, reducing simulation accuracy. Specifically, we found that the UF 3 variant trained on a subset of the training data, in which nitrogen-rich structures were removed, offered vastly better prediction and simulation accuracy than any other variant. By comparing these UF 3 variants, we highlight the nuanced requirements for creating accurate MLIPs, emphasizing the importance of application-specific training data to achieve optimal performance in modeling complex material behaviors.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Gibson, Jason B. [Univ. of Florida, Gainesville, FL (United States)] (ORCID:0000000179745264), Janicki, Tesia Danielle [Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)] (ORCID:0000000160086414), Hire, Ajinkya C. [Univ. of Florida, Gainesville, FL (United States)] (ORCID:0000000331472521), Bishop, Chris [Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)], Lane, J. Matthew Doyle [Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)] (ORCID:0000000278789158), Hennig, Richard G. [Univ. of Florida, Gainesville, FL (United States)] (ORCID:0000000349337686). 2025-04-22. When more data hurts: Optimizing data coverage while mitigating diversity-induced underfitting in an ultrafast machine-learned potential. https://doi.org/10.1103/physrevmaterials.9.043802
Cite the original work for its findings. Save a collection to share your selection of sources.