DASH属性:从基于注意力的动态子结构层次结构中估计原子和分子属性
Marc T Lehner1, Paul Katzberger1, Niels Maeder1
1Department of Chemistry and Applied Biosciences, ETH Zürich, Vladimir-Prelog-Weg 2, 8093 Zürich, Switzerland.
The Journal of chemical physics
|August 15, 2024
概括
我们开发了一种快速,可解释的方法,使用基于注意力的动态亚结构层次结构 (DASH) 树来预测原子性质. 这种方法可以在没有昂贵的计算的情况下实现量子力学准确性,从而实现高效的机器学习功能生成.
科学领域:
- 计算化学计算化学
- 机器学习在化学中的应用
- 量子化学 是一个量子化学.
背景情况:
- 准确预测原子性质对于分子建模至关重要.
- 现有的方法往往需要计算上昂贵的量子力学 (QM) 计算.
- 图形神经网络为化学性质预测提供了一种数据驱动的方法.
研究的目的:
- 为了证明动态基于注意力的亚结构层次结构 (DASH) 树用于预测各种原子性能的实用性.
- 为了证明一个单一的DASH树可以捕获多个属性的特征,避免对属性特定的模型重新训练.
- 为生成分子特征提供QM计算的高效,可解释的替代方案.
主要方法:
- 使用以前开发的DASH树构建方法.
- 杆图表神经网络的注意值来定义属性分配的"规则".
- 应用了DASH树来预测各种原子属性而不需要QM计算.
- 将DASH树,原子性质数据集和波函数公开提供.
主要成果:
- DASH树有效地捕捉了当地的原子环境特征.
- 通过使用相同的DASH树来准确预测多个原子属性.
- 该方法提供了高效率和类似于QM的准确性.
- 该方法通过对子结构的可视化提供可解释的预测.
结论:
- DASH树是预测原子属性的通用和高效工具.
- 这种方法显著降低了与产生机器学习分子特征相关的计算成本.
- DASH树的可解释性增强了对预测的信任和理解.
相关概念视频
Molecular Models
38.1K
Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
38.1K
Predicting Molecular Geometry
34.2K
VSEPR Theory for Determination of Electron Pair Geometries
34.2K
The Aufbau Principle and Hund's Rule
47.1K
To determine the electron configuration for any particular atom, we can build the structures in the order of atomic numbers. Beginning with hydrogen, and continuing across the periods of the periodic table, we add one proton at a time to the nucleus and one electron to the proper subshell until we have described the electron configurations of all the elements. This procedure is called the aufbau principle, from the German word aufbau (“to build up”). Each added electron occupies the...
47.1K
Resonance and Hybrid Structures
16.6K
According to the theory of resonance, if two or more Lewis structures with the same arrangement of atoms can be written for a molecule, ion, or radical, the actual distribution of electrons is an average of that shown by the various Lewis structures.
Resonance Structures and Resonance Hybrids
The Lewis structure of a nitrite anion (NO2−) may actually be drawn in two different ways, distinguished by the locations of the N–O and N=O bonds.
Resonance Structures and Resonance Hybrids
The Lewis structure of a nitrite anion (NO2−) may actually be drawn in two different ways, distinguished by the locations of the N–O and N=O bonds.
16.6K
Inductive Effects on Chemical Shift: Overview
1.1K
The protons in unsubstituted alkanes are strongly shielded with chemical shifts below 1.8 ppm. Methine, methylene, and methyl protons appear at approximately 1.7, 1.2 and 0.7 ppm, while the proton signal from methane appears at 0.23 ppm. An electronegative substituent, such as chlorine, withdraws the electron density from the protons, increasing their chemical shift. Progressive substitution of the hydrogens in methane by chlorine shifts the proton signals increasingly downfield, to 3.05 ppm in...
1.1K
Crystal Field Theory - Tetrahedral and Square Planar Complexes
41.8K
Tetrahedral Complexes
Crystal field theory (CFT) is applicable to molecules in geometries other than octahedral. In octahedral complexes, the lobes of the dx2−y2 and dz2 orbitals point directly at the ligands. For tetrahedral complexes, the d orbitals remain in place, but with only four ligands located between the axes. None of the orbitals points directly at the tetrahedral ligands. However, the dx2−y2 and dz2 orbitals (along the Cartesian axes) overlap with the ligands less than the dxy,...
Crystal field theory (CFT) is applicable to molecules in geometries other than octahedral. In octahedral complexes, the lobes of the dx2−y2 and dz2 orbitals point directly at the ligands. For tetrahedral complexes, the d orbitals remain in place, but with only four ligands located between the axes. None of the orbitals points directly at the tetrahedral ligands. However, the dx2−y2 and dz2 orbitals (along the Cartesian axes) overlap with the ligands less than the dxy,...
41.8K


