首页|用于预测键解离能的图神经网络在数据集间的迁移

用于预测键解离能的图神经网络在数据集间的迁移

扫码查看
基于机器学习的方法如神经网络已经广泛在化学研究中被用于对化学性质的快速估算.生成高精度的机器学习模型需要高质量的数据集.本文从不同的数据集训练了图神经网络,并验证了模型在数据集间的迁移.结果表明跨数据集的模型预测可以给出精度较低,但相关度良好的结果,其中的误差主要来自系统误差.迁移预测所得的值域与训练集的值域高度相关.不同的键型在迁移预测中的误差大小有所区别,其中C-H键一致地体现出最小的迁移误差.
Transferring Graph Neural Network Models for Predicting Bond Dissociation Energy between Datasets
Machine learning(ML)approaches like neural networks have been widely used in chemical researches for fast estimat-ing chemical properties.Generating ML models of good precision requires datasets of high quality,which can be difficult to obtain.In this work,we trained graph neural network(GNN)models from different datasets and ver-ified transferring of the models to oth-er datasets.Our result shows that cross-dataset evaluation can give less accu-rate but still correlative prediction re-sults on different datasets.Errors are mainly due to systematic errors.The value range of pre-diction result is highly related to the range of training set.The precisions of different bonds show different distributions.C-H bond constantly gets the highest precision in the tested bonds.

Machine learningGraph neural networkCross validation

霍姚远、江俊

展开 >

中国科学技术大学合肥微尺度物质科学国家研究中心,合肥 230026

机器学习 图神经网络 模型迁移

国家重点研发计划CAS Project for Young Scientists in Basic ResearchInnovation Program for Quantum Science and Technology国家自然科学基金国家自然科学基金

2018YFA0208603YSBR-0052021ZD03033032202530422033007

2024

化学物理学报(英文版)
中国物理学会

化学物理学报(英文版)

CSTPCDEI
影响因子:0.162
ISSN:1674-0068
年,卷(期):2024.37(1)