模拟局部和一般的量子力学属性,以基于注意力的聚合为基础
David Buterez1, Jon Paul Janet2, Steven J Kiddle3
1Department of Computer Science and Technology, University of Cambridge, Cambridge, CB3 0FD, UK. db804@cam.ac.uk.
Communications chemistry
|November 29, 2023
概括
一种新的基于注意力的聚合方法增强了以原子为中心的神经网络,用于分子性质预测. 这种技术在量子化学任务中比传统的聚合方法提高了准确性.
科学领域:
- 计算化学计算化学
- 机器学习 机器学习
- 量子力学就是量子力学.
背景情况:
- 原子中心神经网络 (ACNN) 是近似分子量子化学性质的最先进技术.
- 当前的ACNN经常使用简单的总和或平均聚合,这可能会限制局部或密集性属性的表示权力.
- 现有的聚合方法可能无法完全捕捉复杂的原子间相互作用,这对于准确的预测至关重要.
研究的目的:
- 为ACNN引入一个可学习的,基于注意力的聚合机制.
- 在化学的深度学习模型中,增强从原子到分子表示的转换.
- 通过更好地建模原子相互作用来提高分子性质的预测准确性.
主要方法:
- 开发了一种新的基于注意力的聚合操作,作为现有方法的取代.
- 将注意力聚合集成到已建立的ACNN架构中,如SchNet和DimeNet++.
- 在不同的数据集,分子性质和理论水平上评估性能.
主要成果:
- 提出的注意力聚合方法始终表现优于总和,平均值和物理意识的聚合方法.
- 取得了显著的性能提升,在特定任务上提高了高达85%.
- 证明了可学习聚合在捕捉复杂的原子相互作用的有效性.
结论:
- 基于注意力的聚合为分子性质预测的传统方法提供了一种优越的替代方案.
- 这种方法增强了ACNN的代表性能力,而不会改变核心架构组件.
- 开发的聚合机制代表了量子化学的几何深度学习的重大进步.
相关概念视频
02:45The Quantum-Mechanical Model of an Atom
42.4K
Shortly after de Broglie published his ideas that the electron in a hydrogen atom could be better thought of as being a circular standing wave instead of a particle moving in quantized circular orbits, Erwin Schrödinger extended de Broglie’s work by deriving what is now known as the Schrödinger equation. When Schrödinger applied his equation to hydrogen-like atoms, he was able to reproduce Bohr’s expression for the energy and, thus, the Rydberg formula governing hydrogen spectra.
42.4K
02:00Molecular Models
38.5K
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.
38.5K
03:51Molecular Orbital Theory II
19.2K
Molecular Orbital Energy Diagrams
19.2K
02:35Molecular Orbital Theory I
32.1K
Overview of Molecular Orbital Theory
32.1K
01:18Structure of Benzene: Molecular Orbital Model
9.1K
According to the molecular orbital (MO) model, benzene has a planar structure with a regular hexagon of six sp2 hybridized carbons. As shown in Figure 1, each carbon is bonded to three other atoms with C–C–C and H–C–C bond angles of 120°. The C–H bond length is 109 pm, and the C–C bond length is 139 pm which is midway between the single bond length of sp3 hybridized carbons (154 pm) and sp2 hybridized carbons (133 pm).
9.1K
01:37Super-resolution Fluorescence Microscopy
7.0K
Super-resolution fluorescence microscopy (SRFM) provides a better resolution than conventional fluorescence microscopy by reducing the point spread function (PSF). PSF is the light intensity distribution from a point that causes it to appear blurred. Due to PSF, each fluorescing point appears bigger than its actual size, and it is the PSF interference of nearby fluorophores that causes the blurred image. Various approaches to achieving higher resolution through SRFM have recently been...
7.0K