E(n)-Equivariant Cartesian Tensor Passing Potential
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910584997412864 |
|---|---|
| author | Wang, Junjie Wang, Yong Zhang, Haoting Yang, Ziyang Liang, Zhixin Shi, Jiuyang Wang, Hui-Tian Xing, Dingyu Sun, Jian |
| author_facet | Wang, Junjie Wang, Yong Zhang, Haoting Yang, Ziyang Liang, Zhixin Shi, Jiuyang Wang, Hui-Tian Xing, Dingyu Sun, Jian |
| contents | Machine learning potential (MLP) has been a popular topic in recent years for its potential to replace expensive first-principles calculations in some large systems. Meanwhile, message passing networks have gained significant attention due to their remarkable accuracy, and a wave of message passing networks based on Cartesian coordinates has emerged. However, the information of the node in these models is limited to scalars, vectors, and tensors. In this work, we proposed High-order Tensor Passing Potential (HotPP), an E(n) equivariant message passing neural network that extends the node embedding and message to an arbitrary order tensor. By performing some basic equivariant operations, high order tensors can be coupled very simply and thus the model can make direct predictions of high-order tensors such as dipole moments and polarizabilities without any modifications. Compared to high order tensor models based on spherical vectors, this network is simpler and can achieve comparable accuracy with much fewer parameters. The tests in several datasets demonstrate HotPP is a promising new approach that warrants further investigation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_15286 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | E(n)-Equivariant Cartesian Tensor Passing Potential Wang, Junjie Wang, Yong Zhang, Haoting Yang, Ziyang Liang, Zhixin Shi, Jiuyang Wang, Hui-Tian Xing, Dingyu Sun, Jian Computational Physics Materials Science J.2 Machine learning potential (MLP) has been a popular topic in recent years for its potential to replace expensive first-principles calculations in some large systems. Meanwhile, message passing networks have gained significant attention due to their remarkable accuracy, and a wave of message passing networks based on Cartesian coordinates has emerged. However, the information of the node in these models is limited to scalars, vectors, and tensors. In this work, we proposed High-order Tensor Passing Potential (HotPP), an E(n) equivariant message passing neural network that extends the node embedding and message to an arbitrary order tensor. By performing some basic equivariant operations, high order tensors can be coupled very simply and thus the model can make direct predictions of high-order tensors such as dipole moments and polarizabilities without any modifications. Compared to high order tensor models based on spherical vectors, this network is simpler and can achieve comparable accuracy with much fewer parameters. The tests in several datasets demonstrate HotPP is a promising new approach that warrants further investigation. |
| title | E(n)-Equivariant Cartesian Tensor Passing Potential |
| topic | Computational Physics Materials Science J.2 |
| url | https://arxiv.org/abs/2402.15286 |