Teacher-student training improves accuracy and efficiency of machine learning interatomic potentials

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Matin, Sakib, Allen, Alice E. A., Shinkle, Emily, Pachalieva, Aleksandra, Craven, Galen T., Nebgen, Benjamin, Smith, Justin S., Messerly, Richard, Li, Ying Wai, Tretiak, Sergei, Barros, Kipton, Lubbers, Nicholas
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916792319868928
author Matin, Sakib
Allen, Alice E. A.
Shinkle, Emily
Pachalieva, Aleksandra
Craven, Galen T.
Nebgen, Benjamin
Smith, Justin S.
Messerly, Richard
Li, Ying Wai
Tretiak, Sergei
Barros, Kipton
Lubbers, Nicholas
author_facet Matin, Sakib
Allen, Alice E. A.
Shinkle, Emily
Pachalieva, Aleksandra
Craven, Galen T.
Nebgen, Benjamin
Smith, Justin S.
Messerly, Richard
Li, Ying Wai
Tretiak, Sergei
Barros, Kipton
Lubbers, Nicholas
contents Machine learning interatomic potentials (MLIPs) are revolutionizing the field of molecular dynamics (MD) simulations. Recent MLIPs have tended towards more complex architectures trained on larger datasets. The resulting increase in computational and memory costs may prohibit the application of these MLIPs to perform large-scale MD simulations. Here, we present a teacher-student training framework in which the latent knowledge from the teacher (atomic energies) is used to augment the students' training. We show that the light-weight student MLIPs have faster MD speeds at a fraction of the memory footprint compared to the teacher models. Remarkably, the student models can even surpass the accuracy of the teachers, even though both are trained on the same quantum chemistry dataset. Our work highlights a practical method for MLIPs to reduce the resources required for large-scale MD simulations.
format Preprint
id arxiv_https___arxiv_org_abs_2502_05379
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Teacher-student training improves accuracy and efficiency of machine learning interatomic potentials
Matin, Sakib
Allen, Alice E. A.
Shinkle, Emily
Pachalieva, Aleksandra
Craven, Galen T.
Nebgen, Benjamin
Smith, Justin S.
Messerly, Richard
Li, Ying Wai
Tretiak, Sergei
Barros, Kipton
Lubbers, Nicholas
Chemical Physics
Machine Learning
Machine learning interatomic potentials (MLIPs) are revolutionizing the field of molecular dynamics (MD) simulations. Recent MLIPs have tended towards more complex architectures trained on larger datasets. The resulting increase in computational and memory costs may prohibit the application of these MLIPs to perform large-scale MD simulations. Here, we present a teacher-student training framework in which the latent knowledge from the teacher (atomic energies) is used to augment the students' training. We show that the light-weight student MLIPs have faster MD speeds at a fraction of the memory footprint compared to the teacher models. Remarkably, the student models can even surpass the accuracy of the teachers, even though both are trained on the same quantum chemistry dataset. Our work highlights a practical method for MLIPs to reduce the resources required for large-scale MD simulations.
title Teacher-student training improves accuracy and efficiency of machine learning interatomic potentials
topic Chemical Physics
Machine Learning
url https://arxiv.org/abs/2502.05379