arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.08026physics.chem-phcs.LG
收藏

OrbNet:使用对称性自适应原子轨道特征的量子化学深度学习

OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features

California Institute of Technology(加州理工学院);Entos, Inc.(Entos公司)

Zhuoran Qiao, Matthew Welborn, Animashree Anandkumar, Frederick R. Manby, Thomas F. Miller

摘要:

我们介绍了一种机器学习方法,其中薛定谔方程的能量解决方案预测使用对称性适应原子轨道的功能和图形神经网络架构。\textsc{OrbNet}被证明优于现有的方法在学习效率和可转移性方面的预测密度泛函理论的结果,同时采用低成本的功能,从半经验的电子结构计算。对于药物样分子的数据集,包括QM 7 b-T,QM 9,GDB-13-T,DrugBank,以及Folmsbee和Hutchison的构象基准数据集的应用,\textsc{OrbNet}预测能量在DFT的化学精度内,计算成本降低了千倍或更多。

英文摘要:

We introduce a machine learning method in which energy solutions from the Schrodinger equation are predicted using symmetry adapted atomic orbitals features and a graph neural-network architecture. \textsc{OrbNet} is shown to outperform existing methods in terms of learning efficiency and transferability for the prediction of density functional theory results while employing low-cost features that are obtained from semi-empirical electronic structure calculations. For applications to datasets of drug-like molecules, including QM7b-T, QM9, GDB-13-T, DrugBank, and the conformer benchmark dataset of Folmsbee and Hutchison, \textsc{OrbNet} predicts energies within chemical accuracy of DFT at a computational cost that is thousand-fold or more reduced.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑