Apax:一个灵活而高性能的框架,用于开发机器学习的原子间潜能
Moritz R Schäfer1, Nico Segreto1, Fabian Zills2
1Institute for Theoretical Chemistry, University of Stuttgart, Pfaffenwaldring 55, 70569 Stuttgart, Germany.
Journal of chemical information and modeling
|July 30, 2025
概括
在JAX (apax) 中的原子学习潜力是一个新的开源软件包,用于机器学习的原子间潜力. 它提供了高效的训练和推断,通过积极学习加速材料发现,并提高了模型准确性.
科学领域:
- 计算材料科学 计算材料科学
- 机器学习用于科学计算.
背景情况:
- 机器学习的原子间潜力 (MLIP) 对于在原子尺度上模拟材料至关重要.
- 现有的MLIP框架可能是计算密集型,并且缺乏灵活性,用于像主动学习这样的高级应用程序.
研究的目的:
- 在JAX (apax) 中介绍Atomistic的学习潜力,这是一款用于MLIPs的开源软件包.
- 展示Apax在加速材料模拟和主动学习工作流程方面的能力.
主要方法:
- 开发了利用JAX框架进行GPU加速和灵活模型设计的Apax.
- 积极学习的实施功能:基于内核的数据选择,不确定性估计和增强的抽样.
- 在离子液体和Li3PO4数据集上训练模型,将性能与现有方法进行比较.
主要成果:
- 在Apax的持续学习模型中,积极学习的培训时间减少了高达85%.
- 在Apax中的高斯时刻神经网络实现了比Allegro更高的准确性和10倍更快的推理速度.
- 在数据并行训练和复杂模型组合的模块化方面表现出良好的可扩展性.
结论:
- apax为开发和部署MLIP提供了一个灵活,高效和用户友好的平台.
- 该软件促进了像主动学习这样的高级应用程序,从而实现更快,更准确的材料模拟.
- apax可以开发新的MLIP模型,并加快材料发现的步伐.
更多相关视频
12:11Computation of Atmospheric Concentrations of Molecular Clusters from ab initio Thermochemistry
Published on: April 8, 2020
8.3K
10:52Multiscale Sampling of a Heterogeneous Water/Metal Catalyst Interface using Density Functional Theory and Force-Field Molecular Dynamics
Published on: April 12, 2019
12.9K
相关概念视频
Atomic Orbitals
35.6K
An atomic orbital represents the three-dimensional regions in an atom where an electron has the highest probability to reside. The radial distribution function indicates the total probability of finding an electron within the thin shell at a distance r from the nucleus. The atomic orbitals have distinct shapes which are determined by l, the angular momentum quantum number. The orbitals are often drawn with a boundary surface, enclosing densest regions of the cloud.
35.6K
Predicting Molecular Geometry
36.1K
VSEPR Theory for Determination of Electron Pair Geometries
36.1K
Electronic Structure of Atoms
24.3K
An atom comprises protons and neutrons, which are contained inside the dense, central core called the nucleus, with electrons present around the nucleus. Taking into account the wave–particle duality of electrons and the uncertainty in position around the nucleus, quantum mechanics provides a more accurate model for the atomic structure. It describes atomic orbitals as the regions around the nucleus where electrons of discrete energy exist, characterized by four quantum...
24.3K
Thermodynamic Potentials
973
Thermodynamic potentials are state functions that are extremely useful in analyzing a thermodynamic system. They have dimensions of energy. The four important thermodynamic potentials are internal energy, enthalpy, Helmholtz free energy, and Gibbs free energy. These thermodynamic potentials can be expressed using two of the following variables: pressure, volume, temperature, and entropy. These two variables are expressed as the rate of change of the thermodynamic potential with respect to other...
973
Force and Potential Energy in One Dimension
5.5K
Force can be calculated from the expression for potential energy, which is a function of position. The component of a conservative force, in a particular direction, equals the negative of the derivative of the corresponding potential energy with respect to the displacement in that direction. For regions where potential energy changes rapidly with displacement, the work done and force is maximum. Also, when force is applied along the positive coordinate axis, the potential energy decreases with...
5.5K
Molecular Orbital Theory I
32.9K
Overview of Molecular Orbital Theory
32.9K
